{"as_of":"2026-08-16T01:46:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:819c1c37e37aff6a5e76e77d2b71532e642e799be0e94a476540178b379bd37d","coverage":[{"denominator":67,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":67,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-06T17:02:44.482544Z","state":"measured"},{"denominator":95,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":95,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-15T06:32:42.880941+00:00","state":"measured"},{"denominator":28,"lane":"inbound_itemization","note":"Pith citing papers itemized under the disclosed page cap.","records_observed":28,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-15T16:52:48.796751Z","state":"measured"},{"denominator":1,"lane":"external_citation_measurements","note":"A source-named dated measurement, never combined with another source.","records_observed":0,"source":"arxiv_reference","source_observed_at":"2026-07-03T22:29:00.387986Z","state":"measured"}],"external_citation_measurements":[],"inbound":[{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-15T16:52:48.796751Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2508.20258","last_updated":"2025-08-27T20:30:43Z","snapshot_observed_at":"2026-08-15T16:46:02.920934Z","submitted_at":"2025-08-27T20:30:43Z","title":"SwizzlePerf: Hardware-Aware LLMs for GPU Kernel Performance Optimization","version":1},"reference_index":8,"source":"pdf_text","source_observed_at":"2026-08-15T16:52:48.796751Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2508.20258"},"observation_digest":"sha256:f7934fdedd176b6e0d7bf1c7618774f86340288eda2e4c8a455c1202395628c7","observation_id":"b683da53-5549-4f81-b824-1b0157d8bf3c","resolution":{"observed_at":"2026-08-15T16:52:48.796751Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-03T19:04:27.222436Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2512.02551","last_updated":"2026-08-05T18:31:31Z","snapshot_observed_at":"2026-08-09T23:09:23.088862Z","submitted_at":"2025-12-02T09:20:15Z","title":"CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication through Reinforcement Learning","version":3},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-08-03T19:04:27.222436Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2512.02551"},"observation_digest":"sha256:32449abd868228da271be17eb60f1b0167a299d9c1d15060e6c4697577cc27e8","observation_id":"535ee78e-d49f-4a8e-94c9-c5b3e57f91fa","resolution":{"observed_at":"2026-08-03T19:04:27.222436Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-03T13:45:23.116995Z","title":"Kevin: Multi-turn rl for generating cuda kernels.arXiv preprint arXiv:2507.11948,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2512.23236","last_updated":"2026-07-06T22:18:07Z","snapshot_observed_at":"2026-08-14T21:09:28.913440Z","submitted_at":"2025-12-29T06:31:55Z","title":"KernelEvolve: Scaling Agentic Kernel Coding for Heterogeneous AI Accelerators at Meta","version":4},"reference_index":2023,"source":"pdf_text","source_observed_at":"2026-08-03T13:45:23.116995Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2512.23236"},"observation_digest":"sha256:ad20eaee84a0657ebe08920816e7656ba262a48d137d205b76ec17efaffe1144","observation_id":"e063c74a-bf05-49cb-808f-89a132b5fac5","resolution":{"observed_at":"2026-08-03T13:45:23.116995Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2603.23566","last_updated":"2026-05-16T02:26:58Z","snapshot_observed_at":"2026-08-13T02:42:19.968149Z","submitted_at":"2026-03-24T08:54:53Z","title":"AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization","version":2},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-05-21T10:16:15.305706Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2603.23566"},"observation_digest":"sha256:a03733cc167f26533bf4bc964dc3aee35abcc7401ffed9665b72eb7f910f6ad0","observation_id":"9fba3d5f-733d-4ee9-99d1-7f48f8b58973","resolution":{"observed_at":"2026-05-21T10:20:00.045508Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2603.28342","last_updated":"2026-04-23T09:14:32Z","snapshot_observed_at":"2026-08-02T23:18:14.887276Z","submitted_at":"2026-03-30T12:12:49Z","title":"Kernel-Smith: A Unified Recipe for Evolutionary Kernel Optimization","version":2},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-05-14T22:17:13.431164Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2603.28342"},"observation_digest":"sha256:279b89487d25c014d838444cc8eaa7edab2c0562c22e5a1d6ffe80aef72ec607","observation_id":"3f03d00a-cb56-4945-9ba7-03aa5b4764fd","resolution":{"observed_at":"2026-05-14T22:18:04.181510Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2604.03144","last_updated":"2026-04-03T16:06:25Z","snapshot_observed_at":"2026-08-11T17:05:18.478668Z","submitted_at":"2026-04-03T16:06:25Z","title":"InCoder-32B-Thinking: Industrial Code World Model for Thinking","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-05-13T18:56:31.975626Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2604.03144"},"observation_digest":"sha256:781b25c82e522784f4258795a0d5f9107ae12c8521d7a060bf673df3ec9b14e1","observation_id":"3faded56-c20c-4182-b3db-3ba1b81f46cd","resolution":{"observed_at":"2026-05-13T18:58:08.660491Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2604.16625","last_updated":"2026-04-17T18:25:03Z","snapshot_observed_at":"2026-08-15T00:42:48.171342Z","submitted_at":"2026-04-17T18:25:03Z","title":"AdaExplore: Failure-Driven Adaptation and Diversity-Preserving Search for Efficient Kernel Generation","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-05-10T08:27:56.130579Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2604.16625"},"observation_digest":"sha256:b213f2f7773e519e03cb640ccdab9c7388cdf3d9f9a58a999c55b55d0ba48ad8","observation_id":"2d0e2179-5344-4719-be68-2fc528e8caac","resolution":{"observed_at":"2026-05-10T09:08:25.936470Z","resolver_source":"arxiv_id","status":"metadata_mismatch"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2605.08455","last_updated":"2026-05-26T07:39:54Z","snapshot_observed_at":"2026-08-11T00:35:00.640352Z","submitted_at":"2026-05-08T20:24:32Z","title":"CUDABeaver: Benchmarking LLM-Based Automated CUDA Debugging","version":1},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-05-12T01:43:06.644640Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2605.08455"},"observation_digest":"sha256:5058f3ada00e7ebbf6ff417dde25f095c4952ddd578d09f5acb15f0d16358c80","observation_id":"ca411864-36f3-4ac9-b7e4-f52c4b3bbef7","resolution":{"observed_at":"2026-05-12T01:46:14.213083Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2605.08455","last_updated":"2026-05-26T07:39:54Z","snapshot_observed_at":"2026-08-11T00:35:00.640352Z","submitted_at":"2026-05-08T20:24:32Z","title":"CUDABeaver: Benchmarking LLM-Based Automated CUDA Debugging","version":2},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-06-30T22:54:20.299335Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2605.08455"},"observation_digest":"sha256:aad60bb3ce26773fa338f7a20cc6af8922c2903168b470984e50f709ded5f8c9","observation_id":"6eeff244-acd2-4e8c-beec-43828bc6bb5d","resolution":{"observed_at":"2026-07-01T13:35:46.950818Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2605.08467","last_updated":"2026-05-08T20:35:27Z","snapshot_observed_at":"2026-08-15T07:44:48.589882Z","submitted_at":"2026-05-08T20:35:27Z","title":"CUDAHercules: Benchmarking Hardware-Aware Expert-level CUDA Optimization for LLMs","version":1},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-05-12T03:24:06.877354Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2605.08467"},"observation_digest":"sha256:2a243371cb38bd9f144fc1c8e2d2c28c8466067312acbb4bde60601d0f079fca","observation_id":"c72586f9-17b9-4895-aa2b-08df83060519","resolution":{"observed_at":"2026-05-12T03:26:19.470835Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2605.23215","last_updated":"2026-05-22T04:19:04Z","snapshot_observed_at":"2026-08-14T03:09:08.088803Z","submitted_at":"2026-05-22T04:19:04Z","title":"FastKernels: Benchmarking GPU Kernel Generation in Production","version":1},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-05-25T05:20:08.724490Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2605.23215"},"observation_digest":"sha256:39d2b873926e7adb4846a4a3a6d7b0f46c86953b888c513aa83fddb07d62f2d0","observation_id":"0a80694b-cb1f-422f-8b56-397975b7356f","resolution":{"observed_at":"2026-05-25T05:20:24.404916Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2605.25954","last_updated":"2026-05-25T15:29:49Z","snapshot_observed_at":"2026-08-01T23:32:04.047479Z","submitted_at":"2026-05-25T15:29:49Z","title":"Step-TP: A Grounded, Step-Level Dataset with Chain-of-Thought Reasoning for LLM-Guided Tensor Program Optimization","version":1},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-06-29T23:16:05.110912Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2605.25954"},"observation_digest":"sha256:439691a2e86e9bd604ef13f162bd0ce6793c70adaf99b7d4fd2015bd82fd8a17","observation_id":"79b45b7d-ce51-42be-9fc4-c64ebd7345fa","resolution":{"observed_at":"2026-06-29T23:24:01.864782Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2605.28213","last_updated":"2026-05-27T09:32:16Z","snapshot_observed_at":"2026-08-01T13:39:26.288705Z","submitted_at":"2026-05-27T09:32:16Z","title":"Learning When to Optimize: Verified Optimization Skills from Expert GPU-Kernel Lineages","version":1},"reference_index":3,"source":"arxiv_source","source_observed_at":"2026-06-29T12:26:23.359699Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2605.28213"},"observation_digest":"sha256:a193a983f2da801d32fdd39f500f30057969589181a8585be598b5cafb2df360","observation_id":"7fa4db39-d853-4e15-9333-61b9aaad65f7","resolution":{"observed_at":"2026-06-29T12:33:24.684781Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2605.29357","last_updated":"2026-05-28T04:55:14Z","snapshot_observed_at":"2026-08-12T23:58:01.216331Z","submitted_at":"2026-05-28T04:55:14Z","title":"PassNet: Scaling Large Language Models for Graph Compiler Pass Generation","version":1},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-06-29T07:53:06.859337Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2605.29357"},"observation_digest":"sha256:5ad7124e90ec9956b17f9098c2e070aaf01163915fc1b5e7d8e9a4fded2b7c4c","observation_id":"f0273d78-45dd-4ddd-a287-1356ef9dc343","resolution":{"observed_at":"2026-06-29T07:53:13.251882Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2605.30359","last_updated":"2026-08-03T01:31:45Z","snapshot_observed_at":"2026-08-06T23:24:42.531300Z","submitted_at":"2026-05-08T03:41:54Z","title":"Kernel Foundry: A Diagnosis-driven Evolutionary Kernel Optimizer with Multi-Experts","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-06-30T23:33:47.477482Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2605.30359"},"observation_digest":"sha256:d25652f169bddfc859a0543a6b37c28494feb2504ff1c21eb284baabfa1f6741","observation_id":"425de9c4-8294-40a6-836d-03a2e901d2a1","resolution":{"observed_at":"2026-06-30T23:35:07.157854Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-04T05:17:04.965673Z","title":"Kevin: Multi-turn rl for generating cuda kernels.arXiv preprint arXiv:2507.11948, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2605.30359","last_updated":"2026-08-03T01:31:45Z","snapshot_observed_at":"2026-08-06T23:24:42.531300Z","submitted_at":"2026-05-08T03:41:54Z","title":"Kernel Foundry: A Diagnosis-driven Evolutionary Kernel Optimizer with Multi-Experts","version":2},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-08-04T05:17:04.965673Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2605.30359"},"observation_digest":"sha256:cfbb72422553a6c2e890734044363cab20445b1a13ab393ddf84a9e118656761","observation_id":"8d22d16c-a45a-4d68-b026-0fe2912237ef","resolution":{"observed_at":"2026-08-04T05:17:04.965673Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2606.17518","last_updated":"2026-06-16T04:54:23Z","snapshot_observed_at":"2026-07-06T23:53:07.149182Z","submitted_at":"2026-06-16T04:54:23Z","title":"SpecGen: Accelerating Agentic Kernel Optimization with Speculative Generation","version":1},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-06-26T23:35:32.175768Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2606.17518"},"observation_digest":"sha256:b78d02e287224fb375402af39c3dd04cf736a6e53521e62a731ad22505ea401e","observation_id":"0043db8c-d393-48c1-9038-9a0667192661","resolution":{"observed_at":"2026-07-03T22:29:00.389666Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2606.30442","last_updated":"2026-06-29T15:16:33Z","snapshot_observed_at":"2026-08-12T18:39:09.263770Z","submitted_at":"2026-06-29T15:16:33Z","title":"The FIL Hypothesis: Inductive Biases Help with Kernel Engineering","version":1},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-06-30T06:06:37.832064Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2606.30442"},"observation_digest":"sha256:dddd0d0ee5196d819faf29050c05ea47d71c7698807655fb305d0473694500a3","observation_id":"f0db2be6-dc3d-4c4e-9131-33c35e2aaf72","resolution":{"observed_at":"2026-06-30T08:24:26.952647Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":"2507.11948","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-03T22:29:00.387986Z","title":"Kevin: Multi-turn rl for generating cuda kernels","venue":null,"work_id":"5adf1cd8-3a29-431d-9a02-7eb2150b195d","year":2025},"citing_paper":{"arxiv_id":"2607.01590","last_updated":"2026-07-03T02:53:16Z","snapshot_observed_at":"2026-08-12T21:57:51.457337Z","submitted_at":"2026-07-02T01:41:34Z","title":"Hawk: Harnessing Hardware-Aware Knowledge for High-Performance NPU Kernel Generation","version":1},"reference_index":26,"source":"pdf_text","source_observed_at":"2026-07-03T15:01:18.911340Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2607.01590"},"observation_digest":"sha256:f7b7b67a9af271d168afef7c1683ccf38e57004f739a56154319a6f216970140","observation_id":"84b02818-876b-42a6-8020-54e257458065","resolution":{"observed_at":"2026-07-03T15:08:32.483938Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-07-12T08:42:42.063339Z","title":"Kevin: Multi-turn rl for generating cuda kernels,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2607.01590","last_updated":"2026-07-03T02:53:16Z","snapshot_observed_at":"2026-08-12T21:57:51.457337Z","submitted_at":"2026-07-02T01:41:34Z","title":"Hawk: Harnessing Hardware-Aware Knowledge for High-Performance NPU Kernel Generation","version":2},"reference_index":26,"source":"pdf_text","source_observed_at":"2026-07-12T08:42:42.063339Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2607.01590"},"observation_digest":"sha256:cad6088c2851c5c42ed85c91812289cab8051c71a2fd6d9975a80a05bc0e67a3","observation_id":"5cacdf76-dff1-4988-88f2-57f624e95e67","resolution":{"observed_at":"2026-07-12T08:42:42.063339Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-01T15:52:50.146953Z","title":"Carlo Baronio, Pietro Marsella, Ben Pan, Simon Cherniavskii, Janice Wang, and Fotis Iliopoulos","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2607.18171","last_updated":"2026-08-10T06:51:12Z","snapshot_observed_at":"2026-08-13T23:40:18.315443Z","submitted_at":"2026-07-20T17:12:28Z","title":"FlashRT: Agent Harness for Guiding Agents to Deploy Real-Time Multimodal Applications","version":1},"reference_index":2025,"source":"pdf_text","source_observed_at":"2026-08-01T15:52:50.146953Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2607.18171"},"observation_digest":"sha256:448426fd0ee515d1eaecfbe55965b378d12e5a3b6ce3986c9f32df21febaf794","observation_id":"c8e838f2-032c-4fd8-b861-5e53725559d9","resolution":{"observed_at":"2026-08-01T15:52:50.146953Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-02T08:11:50.753543Z","title":"Kevin: Multi-turn RL for generating CUDA kernels.arXiv preprint arXiv:2507.11948, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2607.20518","last_updated":"2026-07-08T02:22:22Z","snapshot_observed_at":"2026-08-14T17:25:19.834750Z","submitted_at":"2026-07-08T02:22:22Z","title":"CANN Bench: Benchmarking Agent Generated Kernels against Real NPU and Algorithmic Limits","version":1},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-08-02T08:11:50.753543Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2607.20518"},"observation_digest":"sha256:b37bfbe1c5ade056ce4367da343cd7da85077421e45c6f74e91d72398fe8e5a7","observation_id":"c7930720-6f3c-4bcc-bd0a-d0a1e1c2176e","resolution":{"observed_at":"2026-08-02T08:11:50.753543Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-15T15:35:43.831799Z","title":null,"venue":null,"work_id":null,"year":2026},"citing_paper":{"arxiv_id":"2607.20908","last_updated":"2026-07-23T04:06:41Z","snapshot_observed_at":"2026-08-15T22:56:18.949146Z","submitted_at":"2026-07-23T04:06:41Z","title":"Multi-turn RL with Structural and Performance Aware Rewards for CUDA Kernel Generation","version":1},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-08-15T15:35:43.831799Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2607.20908"},"observation_digest":"sha256:138dbfcd9bdfdbf6576d2b102fcca52032a2d3f5597a573b10a095810bd4834c","observation_id":"e28fdee3-c240-4fc2-9a30-b7015b90c302","resolution":{"observed_at":"2026-08-15T15:35:43.831799Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-01T03:42:32.030503Z","title":"Kevin: Multi-turn RL for generating CUDA kernels, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2607.23089","last_updated":"2026-07-25T07:49:38Z","snapshot_observed_at":"2026-08-05T17:35:24.794954Z","submitted_at":"2026-07-25T07:49:38Z","title":"Compiler-Grounded Hierarchical Diagnosis for LLM-Based Triton Kernel Optimization","version":1},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-08-01T03:42:32.030503Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2607.23089"},"observation_digest":"sha256:79a61dee15cfbd386067009ec56e80ac1214956ea10e688348f1dfc29210595b","observation_id":"d769efeb-4aa7-474b-a42f-033ec73217b6","resolution":{"observed_at":"2026-08-01T03:42:32.030503Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-01T10:53:44.502494Z","title":"Kevin: Multi-turn RL for generating CUDA kernels.CoRR, abs/2507.11948, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2607.27231","last_updated":"2026-07-22T12:40:11Z","snapshot_observed_at":"2026-08-13T23:11:38.674337Z","submitted_at":"2026-07-22T12:40:11Z","title":"KernelGenBench: A Multi-Source and Multi-Chip Benchmark for LLM-based Kernel Generation","version":1},"reference_index":15,"source":"pdf_text","source_observed_at":"2026-08-01T10:53:44.502494Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2607.27231"},"observation_digest":"sha256:db558898b63d1c435a3d15c44923dfc8fd7865e431897a772e866285e77fe592","observation_id":"4d30955d-56b6-4ef1-aefe-c3be1ffca171","resolution":{"observed_at":"2026-08-01T10:53:44.502494Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-01T10:53:44.585663Z","title":null,"venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2607.27231","last_updated":"2026-07-22T12:40:11Z","snapshot_observed_at":"2026-08-13T23:11:38.674337Z","submitted_at":"2026-07-22T12:40:11Z","title":"KernelGenBench: A Multi-Source and Multi-Chip Benchmark for LLM-based Kernel Generation","version":1},"reference_index":18,"source":"pdf_text","source_observed_at":"2026-08-01T10:53:44.585663Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2607.27231"},"observation_digest":"sha256:6147a03e74a8377d2c830eead5cc40b2f2bd7e4ee62b2e5c9310b93f7ec0fcaa","observation_id":"147af661-1cf0-4198-bd16-8bee0bd98adf","resolution":{"observed_at":"2026-08-01T10:53:44.585663Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-04T04:39:20.468386Z","title":"Kevin: Multi-Turn Reinforcement Learning for Writing CUDA Kernels,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2608.02569","last_updated":"2026-08-03T17:45:58Z","snapshot_observed_at":"2026-08-08T15:31:19.552043Z","submitted_at":"2026-08-03T17:45:58Z","title":"AtumAI: A Principled Framework for Agentic Generation of Datacenter Control-Plane Policies","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-08-04T04:39:20.468386Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2608.02569"},"observation_digest":"sha256:3b0f744088274b1c0734f4c910cf68313d3a12b3a3bda4cc2dcdb5152c01c35d","observation_id":"eaf5dee3-c334-4d6b-ae4b-a95509097072","resolution":{"observed_at":"2026-08-04T04:39:20.468386Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.11948","snapshot_observed_at":"2026-08-07T00:14:20.253791Z","title":null,"venue":null,"work_id":null,"year":2026},"citing_paper":{"arxiv_id":"2608.02712","last_updated":"2026-08-03T17:59:48Z","snapshot_observed_at":"2026-08-15T12:05:45.326571Z","submitted_at":"2026-08-03T17:59:48Z","title":"Don't Regenerate, Debug: A Domain-Specific Agent for Repairing Near-Miss Hardware Operators","version":1},"reference_index":2025,"source":"pdf_text","source_observed_at":"2026-08-07T00:14:20.253791Z"},"links":{"cited_paper":"/paper/2507.11948","citing_paper":"/paper/2608.02712"},"observation_digest":"sha256:722b56143c750cce400adda0259b331838632a7b319f03421fc850e435ae7630","observation_id":"15b11f46-69ff-44f9-ad14-dc34a5344dc9","resolution":{"observed_at":"2026-08-07T00:14:20.253791Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"links":{"evidence":"/evidence","html":"/paper/2507.11948/citation-record","integrity":"/paper/2507.11948/integrity","json":"/paper/2507.11948/citation-record.json","paper":"/paper/2507.11948"},"outbound":[{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:37.716853Z","title":"Concrete problems in ai safety, 2016","venue":null,"work_id":null,"year":2016},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:37.716853Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:9c2adc5601affe839cd435effd920f96315f91bf19c449ba241b89ca2297b67c","observation_id":"0a7bd183-13a3-4e14-b694-791e652ffd31","resolution":{"observed_at":"2026-08-06T17:02:37.716853Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:37.770626Z","title":"Le, Christopher Ré, and Azalia Mirhoseini","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:37.770626Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:242d749225893698d72d02c264eb8ba618f0985c050fb0abc61766492ab74ce9","observation_id":"a5e2c076-e17e-49c8-86eb-fd0b5b618a08","resolution":{"observed_at":"2026-08-06T17:02:37.770626Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:37.908432Z","title":"Gonzalez, and Ion Stoica","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:37.908432Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:454e86f85e81b851f7baf16a117033f0956f5e5c945f8647ba302f933cd86f9d","observation_id":"908ad037-323e-4663-a0c3-90e196941179","resolution":{"observed_at":"2026-08-06T17:02:37.908432Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:38.046691Z","title":null,"venue":null,"work_id":null,"year":2021},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:38.046691Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:5f80a38a77837e0f6863060927a651e636c2430f806db75ef19b07c7b014a59a","observation_id":"264218f1-5fe7-4531-8718-759822c4f075","resolution":{"observed_at":"2026-08-06T17:02:38.046691Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:50.418893Z","title":"Automating gpu kernel generation with deepseek- r1 and inference-time scaling","venue":null,"work_id":"20a4a541-4cb2-494e-9e33-70aa867578f8","year":null},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:38.147054Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:c31bb1ab115803efba4018cd2b196be7ea0929589fb7e7b242e50f2aae69dd7e","observation_id":"486c0110-96a0-423a-97e7-5752f0e22b55","resolution":{"observed_at":"2026-08-06T17:02:50.570501Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:50.164271Z","title":"Warpdrive: An agentic workflow for ninja gpu transformations","venue":null,"work_id":"f5dad10f-d76d-4988-b7e9-7f48891146c2","year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:38.300229Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:a58e8394115b4535abb9e2e38625bb292ff90be8db6c097ca188cca2100db083","observation_id":"7c94f462-af32-41d0-87f2-7a6e94890428","resolution":{"observed_at":"2026-08-06T17:02:50.231647Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:38.412758Z","title":"Flashattention-2: Faster attention with better parallelism and work partitioning, 2023","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:38.412758Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:1e700fd6f56200e7be2589ea95047c84111391d99f5093f32f506dddce29ff41","observation_id":"370f31e3-e3a4-4378-b821-34877dea68ac","resolution":{"observed_at":"2026-08-06T17:02:38.412758Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:38.515282Z","title":"Fu, Stefano Ermon, Atri Rudra, and Christopher Ré","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":8,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:38.515282Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:f257d272c5ee2dc35c69f1042ec78e80472bff4886f52fc5bb307d58b840a727","observation_id":"8401544f-6539-406d-adea-773b174f9047","resolution":{"observed_at":"2026-08-06T17:02:38.515282Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:38.628308Z","title":"Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":9,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:38.628308Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:bf2aa15a82ff7be5f43a8a8ce399d2824b7346a06a857d8ff4fd0a3a35099c74","observation_id":"78d9d8b4-9497-48f0-8e98-d64f1ed5b6d4","resolution":{"observed_at":"2026-08-06T17:02:38.628308Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:49.967412Z","title":"Gemini 2.5: Our most intelligent models are getting even bet- ter","venue":null,"work_id":"5263d374-e0f0-4239-94d0-17549b82108a","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:38.711064Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:a9ca5db16eb10b05a288357abc188115feae4b28df73a3257ccebf7c77f15f3b","observation_id":"5b4641b1-e88f-4149-96af-a34f9200e92b","resolution":{"observed_at":"2026-08-06T17:02:50.044013Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:38.800696Z","title":"Bartlett, Ilya Sutskever, and Pieter Abbeel","venue":null,"work_id":null,"year":2016},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:38.800696Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:475b08bacbe505ea757dbe862ac474558084b1f30c3453218d4774cc8e7d8df1","observation_id":"6cf5fccd-df10-476c-89eb-917149a0dab0","resolution":{"observed_at":"2026-08-06T17:02:38.800696Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:38.917350Z","title":"Codemonkeys: Scaling test-time compute for software engineering, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:38.917350Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:c64842e4d84d96c0d35979891e52fbfd00b8e39d314622bb0ad27eb49ea16faa","observation_id":"5c0292a5-2658-44dd-abb4-83e6b2b5e678","resolution":{"observed_at":"2026-08-06T17:02:38.917350Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:39.032733Z","title":"Rlef: Grounding code llms in execution feedback with reinforcement learning, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":13,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:39.032733Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:780d02cfedd8708c46c960bf033210d3e074c8d5ef3859ef2d90e317a6f25ead","observation_id":"dbfa5d30-c513-4a6d-850d-a4290c812794","resolution":{"observed_at":"2026-08-06T17:02:39.032733Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:49.788453Z","title":null,"venue":null,"work_id":"c3feb354-44d7-4b44-a20b-3789d2bd1259","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":14,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:39.123994Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:4b63d5ff281741be9ae3f049a4e25f8e8afa72cd925e7f6eb0362a8f3e18ad00","observation_id":"ab71fa39-8595-4a41-8c0e-3d490ebba9cc","resolution":{"observed_at":"2026-08-06T17:02:49.851401Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:49.653539Z","title":"Alphaevolve: A gemini-powered coding agent for designing advanced algorithms, May 2025","venue":null,"work_id":"5c1228cd-7c2d-42f1-af1c-a8835962a52a","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":15,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:39.223028Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:248a823ca88881c9ed94a417a1bfe8a658aa45cc52c3ec3103c0e493babb9a6d","observation_id":"b2f20324-459e-41a5-a330-a0267c0dfe81","resolution":{"observed_at":"2026-08-06T17:02:49.722393Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:49.557156Z","title":"Openrlhf: An easy-to-use, scalable and high-performance rlhf framework, 2024","venue":null,"work_id":"41cbc021-fea8-4bf0-8ae4-8438031d3580","year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":16,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:39.280864Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:19b97447765c43b91641720c9ae539f22296069431263c224b0dfbc171781081","observation_id":"adda0869-c854-408e-ac5a-d58827759369","resolution":{"observed_at":"2026-08-06T17:02:49.573632Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:39.365559Z","title":"Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik Narasimhan","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:39.365559Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:f6647aa4674c6af7eaef7220d646b5408e99b008e02e0ca908d707a22720de9c","observation_id":"9c017636-41aa-43bb-bf7f-b490aa865b74","resolution":{"observed_at":"2026-08-06T17:02:39.365559Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:39.454227Z","title":"The stack: 3 tb of permissively licensed source code, 2022","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":18,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:39.454227Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:21ecef008de1722d28c22cbef813b2fce8d328f5cd7dd29aa68e43a66dcf38f4","observation_id":"3f538d22-e0cd-467f-ad65-b57332470382","resolution":{"observed_at":"2026-08-06T17:02:39.454227Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:49.379951Z","title":"Gonzalez, Hao Zhang, and Ion Stoica","venue":null,"work_id":"b56dc609-0881-4968-9b7e-f859eea382d0","year":2023},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":19,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:39.579448Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:ac5ac59139fe1aa17999dcc11b85ccdf05d4833a86f484704ee3eb2c702a1d62","observation_id":"994d721d-8ef2-4570-881c-fe7542d6400a","resolution":{"observed_at":"2026-08-06T17:02:49.420206Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:39.670738Z","title":"Miranda, Alisa Liu, Nouha Dziri, Shane Lyu, Yuling Gu, Saumya Malik, Victoria Graf, Jena D","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":20,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:39.670738Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:ecf67e34ab989ac84e3e82d2e7cbf3f360119cb6ee91abaca18385be58eadea4","observation_id":"985963c8-f033-409a-920e-90e66ba15741","resolution":{"observed_at":"2026-08-06T17:02:39.670738Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:49.196993Z","title":"The ai cuda engineer: Agentic cuda kernel discovery, optimization and composition, 2025","venue":null,"work_id":"402d5d73-2d7e-4e73-9f8d-5767035dda22","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":21,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:39.810647Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:72c30c2318270744ee09c5be6128b60b72121f860ed080e4432522a4dc32629e","observation_id":"ea6ec814-cef0-4f46-b399-a188e62a4ca1","resolution":{"observed_at":"2026-08-06T17:02:49.252860Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:39.889019Z","title":"Tritonbench: Benchmarking large language model capabilities for generating triton operators, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":22,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:39.889019Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:e35c17cc59c43b8c542177ed10f9c12aca518414af6a43f383785b4f1da25e14","observation_id":"fbbd5c5c-1d72-40ee-8b94-5cbd3a3a7355","resolution":{"observed_at":"2026-08-06T17:02:39.889019Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:49.039624Z","title":"Starcoder: may the source be with you!, 2023","venue":null,"work_id":"61ddd03d-ff75-45d8-bb1b-619c062250ea","year":2023},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":23,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:40.058883Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:e97b4109df7eac5f9d6aecaa286f8f0f2bc121ae8f8d897b3aaa9cfa16c1b8f5","observation_id":"9f951b42-dbe5-4b0d-9d1d-f0f4e15e132c","resolution":{"observed_at":"2026-08-06T17:02:49.082607Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:48.853834Z","title":"Rltf: Reinforcement learning from unit test feedback, 2023","venue":null,"work_id":"5f6819d9-47c7-43fc-b616-a87128200720","year":2023},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":24,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:40.131633Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:89d88c6d7119cea6fc1b9838379540b5c513d2535403f7552c4fd0774f4f9347","observation_id":"a8f83dc6-396b-4c8b-ae39-af5869b70f18","resolution":{"observed_at":"2026-08-06T17:02:48.934917Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:40.230633Z","title":"Understanding r1-zero-like training: A critical perspective, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":25,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:40.230633Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:aaac999ff60d89baf3fbe3a588b97b35e3b71d54f1288f83080e0ab63155d9e4","observation_id":"6e3c9e77-8c31-4a7d-a15c-b33dd2abef31","resolution":{"observed_at":"2026-08-06T17:02:40.230633Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:48.630636Z","title":"Deepcoder: A fully open-source 14b coder at o3-mini level","venue":null,"work_id":"c04e4e96-e4f1-417d-b7f1-1703b5d07f25","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":26,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:40.319204Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:df3de3607844b7a5afa9d7c495d4f1e699802857f1e191af56ab9ddcd301d46b","observation_id":"8a78a428-35c7-487f-93e8-6dac163f1c63","resolution":{"observed_at":"2026-08-06T17:02:48.735720Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:48.475871Z","title":"Tang, Manan Roongta, Colin Cai, Jeffrey Luo, Li Erran Li, Raluca Ada Popa, and Ion Sto- ica","venue":null,"work_id":"477f8d9c-c088-4e5c-b185-8b200233a66b","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":27,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:40.392131Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:ee1eba66f83f3b5f48e4ea6d1d7d86b4163400fe9a57264076bae47eb64d8be1","observation_id":"3820b2c3-5891-45c4-a740-bead6bf69795","resolution":{"observed_at":"2026-08-06T17:02:48.526909Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:48.320061Z","title":"Measuring automated kernel engineering, February 2025","venue":null,"work_id":"188e0f6e-1f7b-4aef-b77e-344f555197ac","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":28,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:40.479226Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:53840d8a2c6d87a88e62c70660ce18ebe265b22263e14c185b37d721317b18ed","observation_id":"5142138f-e7d3-483c-83a3-2f56393ed435","resolution":{"observed_at":"2026-08-06T17:02:48.385886Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:48.191263Z","title":"Scalable parallel programming with cuda","venue":null,"work_id":"fa115057-61ac-4b20-8c74-c6d28998be72","year":2008},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":29,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:40.634379Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:91b900c441081534f374df4ac589d35ca0f4f32b8948f478c7ca319afe1ec9ce","observation_id":"19c8e387-3eb3-4479-b82a-d23999ecca0f","resolution":{"observed_at":"2026-08-06T17:02:48.230006Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:48.059745Z","title":"Lee, Ed H","venue":null,"work_id":"163fc064-8b6a-4cc8-bf56-8a8998879229","year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":30,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:40.744190Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:ba95ddf9b1895be8826ef9f564d985e2c7a8b3c3f9091b121822a332e7fa1ed6","observation_id":"599a9aa2-242e-49ba-a9ca-591ec0f3ee9b","resolution":{"observed_at":"2026-08-06T17:02:48.128205Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:47.926364Z","title":"Gpu mode at nvidia gtc 2025, 2025","venue":null,"work_id":"8d1d8698-0cdd-48e5-a8e2-802b7ddf1cb9","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":31,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:40.821890Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:e9acbab5d95a05495253a926e75de522c8402b359212c20b21cb4be388affcfe","observation_id":"7fcfb524-b4f7-456c-8370-c5e3bb0b5b5b","resolution":{"observed_at":"2026-08-06T17:02:47.998420Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:47.805656Z","title":"Cutlass: Cuda templates for linear algebra subroutines, May 2025","venue":null,"work_id":"ac35fd98-a421-4bb2-8d2b-4b2e67004744","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":32,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:40.922902Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:3a05e18b463c280a310f42c227297daf8cb38773b8906a03f543c09eeaea3d25","observation_id":"b43bf19a-f50d-4911-a7f0-2b96b7b57986","resolution":{"observed_at":"2026-08-06T17:02:47.862699Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:41.007723Z","title":"Zhang, William Hu, Christopher Ré, and Azalia Mirhoseini","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":33,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:41.007723Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:682ceccca222d41cbaff31eb135b37350484b8cc8d516da235556be89123884e","observation_id":"1828f03e-cde0-45e5-b15f-075b54006f98","resolution":{"observed_at":"2026-08-06T17:02:41.007723Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:47.659685Z","title":null,"venue":null,"work_id":"df3dc3d9-7d30-48b6-8d38-07cb7648dfaf","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":34,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:41.166521Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:8fc7c3c673722a7dd10af01966a76831b3d162969f7b277f5d7c38cae07d1f04","observation_id":"ec612b63-b133-419e-a852-7e0a9d80dd9c","resolution":{"observed_at":"2026-08-06T17:02:47.740822Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:47.512239Z","title":"Putting the value back in rl: Better test-time scaling by unifying llm reasoners with verifiers, 2025","venue":null,"work_id":"8dc9818b-0f61-4d28-aea6-ce27e0148f7f","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":35,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:41.287739Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:d5e1443f6b41eaae4f0749650a174a8d82b39cc2ac7841a66014709a1bcb6634","observation_id":"b6e797ff-9c65-41be-95bb-9919601a1ceb","resolution":{"observed_at":"2026-08-06T17:02:47.585512Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:41.365977Z","title":"Llms are greedy agents: Effects of rl fine-tuning on decision-making abilities, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":36,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:41.365977Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:6adc409fe22cc61873165b2cffcbfd2b4b1664253148ea2a5c0e228ba0de2bfc","observation_id":"4fe2a9fa-629e-45e7-82f7-4229c247c509","resolution":{"observed_at":"2026-08-06T17:02:41.365977Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:41.426044Z","title":"Proximal policy optimization algorithms, 2017","venue":null,"work_id":null,"year":2017},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":37,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:41.426044Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:f1d9e73969206f9f164407d28d102404d55765c22d147e222404038d0171ce5a","observation_id":"c836b34c-cedd-4407-a84a-3929f95b46ea","resolution":{"observed_at":"2026-08-06T17:02:41.426044Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:47.368108Z","title":null,"venue":null,"work_id":"c63a51be-ec25-41f2-88d3-e4042963ec9d","year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":38,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:41.522138Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:5117bd4baff07575701917c6af2d71d38defb0ef2a271ed51b9446b07a103024","observation_id":"b6d68860-f949-4439-87f7-0b5e5991ae9d","resolution":{"observed_at":"2026-08-06T17:02:47.412569Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:41.621229Z","title":"Learning performance-improving code edits, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":39,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:41.621229Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:dea277c11fc558512dcfb42c95df76fde661e9a88d8f741ab65ce559f665dfcb","observation_id":"e8f066b2-7645-4f43-8b42-b21236371ce2","resolution":{"observed_at":"2026-08-06T17:02:41.621229Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:41.736142Z","title":"Mastering chess and shogi by self-play with a general reinforcement learning algorithm, 2017","venue":null,"work_id":null,"year":2017},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":40,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:41.736142Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:f446a3969f7beb714784aadbbd9b895c7134378067ecb5c62ad9762e227a3bae","observation_id":"c6f27d15-1239-441f-ba9f-2b284b44abf1","resolution":{"observed_at":"2026-08-06T17:02:41.736142Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:41.844789Z","title":"Scaling llm test-time compute optimally can be more effective than scaling model parameters, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":41,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:41.844789Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:1a602884e04136ac8e2c89c52e7d665dea4294c025f2ca346f3c45d8ff2ece1e","observation_id":"45a75b80-3026-4195-8576-16949a5ca949","resolution":{"observed_at":"2026-08-06T17:02:41.844789Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:47.220033Z","title":"Spector, Simran Arora, Aaryan Singhal, Daniel Y","venue":null,"work_id":"9076abb1-e625-43a5-90ae-f920c15f243c","year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":42,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:41.933987Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:22b3fbea97c6e81d934d22c1a2fc6a72b028c602c5a0238e8c3426e5337fb527","observation_id":"ae984368-78e2-494f-99aa-c2b3f3ea0a5c","resolution":{"observed_at":"2026-08-06T17:02:47.274098Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:47.115404Z","title":"Training a generally curious agent, 2025","venue":null,"work_id":"e8138471-3611-4a26-a01b-f842c38dbe25","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":43,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:42.033903Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:5d7907f8928fa53d0d16b3fdaf7ee67d44762d5eb12cc9e78a8f07e1920da5dc","observation_id":"83cd90d4-c30d-42ab-b147-ef16c01e9f67","resolution":{"observed_at":"2026-08-06T17:02:47.174414Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:42.120434Z","title":"Kimi k1.5: Scaling reinforcement learning with llms, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":44,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:42.120434Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:ac728b9ca0917f0cf35381ad4beba996499b9a3445a6ee1f1b5c97210441c696","observation_id":"af9ae08a-7ba4-467b-b614-cac24938be24","resolution":{"observed_at":"2026-08-06T17:02:42.120434Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:42.215393Z","title":"Sky-t1: Train your own o1 preview model within $450","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":45,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:42.215393Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:8785dd84af70d179d100931f79aa518cf2ec4bf416e2986d706449fa12e448ea","observation_id":"42140f02-9784-4ffb-bb11-b7c00a515fcc","resolution":{"observed_at":"2026-08-06T17:02:42.215393Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:46.955075Z","title":"Intellect-2: A reasoning model trained through globally decentralized reinforcement learning, 2025","venue":null,"work_id":"d8fd6b28-416a-4206-ab4a-a032cb4d12da","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":46,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:42.329149Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:6645b250eebcbf583a30f99f1d8c32f92b28fbdd8342a84d8b034b0a73071d24","observation_id":"1ee8bf8e-6f57-4746-b41d-ee73f2e5e71a","resolution":{"observed_at":"2026-08-06T17:02:46.993805Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:42.457168Z","title":"Qwq-32b: Embracing the power of reinforcement learning, March 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":47,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:42.457168Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:ef20592b88c0900e22a5469f673507259443f5fe0b807365828e27b5b5a0de68","observation_id":"ca789807-ab56-4df6-be1c-6b82bc6c5408","resolution":{"observed_at":"2026-08-06T17:02:42.457168Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:42.564921Z","title":null,"venue":null,"work_id":null,"year":2019},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":48,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:42.564921Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:ffd5d3fc36d391d5bfa3ace25db8b1dd9e8da06198342d96a644dd025796a074","observation_id":"4bdd59a0-a8a3-4cb3-8f8c-3d9386aa5ea1","resolution":{"observed_at":"2026-08-06T17:02:42.564921Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:46.769075Z","title":"Ecco: Can we improve model-generated code efficiency without sacrificing functional correctness?, 2024","venue":null,"work_id":"4e3050b3-0d7e-42bc-b68c-0544b1b06d85","year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":49,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:42.665133Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:d97f30cf4356791d399098bb2156d82eebd8f8a3273fb9f26867db7b7d6cfa24","observation_id":"6e583a8a-9b63-47b9-8029-da0fd197eec4","resolution":{"observed_at":"2026-08-06T17:02:46.851958Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:46.621383Z","title":"Zero++: Extremely efficient collective commu- nication for giant model training, 2023","venue":null,"work_id":"9023a15e-b6c9-4031-bfb4-d7d63f53389d","year":2023},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":50,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:42.761210Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:a2916dbd31ffc1ee0d622b83f37ddc4e5a4c7aa5b6ce82e543da998ed74c46b9","observation_id":"34c42493-cd15-4ec1-b0e9-491bb6ba8972","resolution":{"observed_at":"2026-08-06T17:02:46.680216Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:42.882939Z","title":"Xu, Xiangru Tang, Mingchen Zhuge, Jiayi Pan, Yueqi Song, Bowen Li, Jaskirat Singh, Hoang H","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":51,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:42.882939Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:05995776b958dd1fb8163a57007be8968e67a6f8db107c9a7b2593859c121455","observation_id":"162d1052-4331-4ebf-953e-d861458fc221","resolution":{"observed_at":"2026-08-06T17:02:42.882939Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:46.475526Z","title":"Reinforcement learning for reasoning in large language models with one training example, 2025","venue":null,"work_id":"1f1b2934-18d5-49aa-90b1-d9241bfa7348","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":52,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:43.027151Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:38e4b5a33e46aac0ea73c0c21aef4a792be3d1f92b705262b50c6cfe61fc4e14","observation_id":"a2a60b80-3d23-4caf-8cba-0e6c4d649727","resolution":{"observed_at":"2026-08-06T17:02:46.531676Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:46.339237Z","title":"Ragen: Understanding self-evolution in llm agents via multi-turn reinforcement learning, 2025","venue":null,"work_id":"9765f258-f5a9-46d7-9174-9ad6b3d8c613","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":53,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:43.209627Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:257e9e1d19f5f631c712d2d3b5dc66b63027af47182fa889817966008924b8c4","observation_id":"a0624456-b9ae-4fac-a8f9-a2be7bc54a56","resolution":{"observed_at":"2026-08-06T17:02:46.400868Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:46.184764Z","title":null,"venue":null,"work_id":"9c16251c-69ff-4562-a96a-aec4caf4cc21","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":54,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:43.323381Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:d254c25a592f2b8786651daaeeda0e6271e0f60741a15549edef17f2daa8bcb7","observation_id":"a07e7628-9de1-4bd5-8bb2-8e0e52a459ec","resolution":{"observed_at":"2026-08-06T17:02:46.243865Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:46.027654Z","title":"Towards system 2 reasoning in llms: Learning how to think with meta chain-of-thought, 2025","venue":null,"work_id":"a0b1462f-371c-4db0-a642-57562ca0cb50","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":55,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:43.416387Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:11dab828fdb5df330967aa50e0c944a2c16b82ae12176777b66cd0c4a7591a19","observation_id":"88cdb48e-9a3f-436e-8e8f-82d24cdb4792","resolution":{"observed_at":"2026-08-06T17:02:46.083782Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2501.01005","last_updated":"2025-04-21T20:10:11Z","snapshot_observed_at":"2026-07-06T20:15:36.280948Z","submitted_at":"2025-01-02T02:02:20Z","title":"FlashInfer: Efficient and Customizable Attention Engine for LLM Inference Serving","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2501.01005","snapshot_observed_at":"2026-08-06T17:02:43.507686Z","title":"Flashinfer: Efficient and customizable attention engine for llm inference serving","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":56,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:43.507686Z"},"links":{"cited_paper":"/paper/2501.01005","citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:64792e5d45847f7f9255c1492ac598458cdeccca3216d11a65cba8b819b0b59f","observation_id":"89100833-2cca-4c22-811e-b7f46327ee73","resolution":{"observed_at":"2026-08-06T17:02:43.507686Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:43.639280Z","title":"Dapo: An open-source llm reinforcement learning system at scale, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":57,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:43.639280Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:ccda99f2c77b9eb482b38391dcabe646c1681a0bcb930f6e5f4dc105f8b15b24","observation_id":"edd9555f-c826-4668-aadc-242e257794aa","resolution":{"observed_at":"2026-08-06T17:02:43.639280Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:43.732164Z","title":"Does reinforcement learning really incentivize reasoning capacity in llms beyond the base model?, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":58,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:43.732164Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:3a4632668e3157ea9f5d8ca8e6b537c594b025534d43ee902b46274cef04e871","observation_id":"c4f35d49-41f7-40be-af7c-a17f4a6accca","resolution":{"observed_at":"2026-08-06T17:02:43.732164Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:45.771808Z","title":"Deepgemm: clean and efficient fp8 gemm kernels with fine-grained scaling","venue":null,"work_id":"c2aab000-ece1-4f7d-8947-c7feb62ad47b","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":59,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:43.849821Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:0d0487b1579d97d2151b1b18caf308835febc5ff961d34931ba7b18a525b2430","observation_id":"2405bc32-944f-4c1c-9e44-e8e1880a011f","resolution":{"observed_at":"2026-08-06T17:02:45.925455Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:43.932409Z","title":"Minif2f: a cross-system benchmark for formal olympiad-level mathematics, 2022","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":60,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:43.932409Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:cebc6e390c3bc37bfda78c0243063fdc4d76d488e2fc473eb7550eae10123cfa","observation_id":"3f4041b6-18c6-4acc-b75a-4324cafd5f7e","resolution":{"observed_at":"2026-08-06T17:02:43.932409Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:45.575258Z","title":"Archer: Training language model agents via hierarchical multi-turn rl, 2024","venue":null,"work_id":"a1120fea-3d6c-4251-8629-cf8ff8284c76","year":2024},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":61,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:44.021267Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:a2b60af8db02ff597a136e30f49072e8b9b0b3658cbe90710b1b266d91c3e6ad","observation_id":"b1af03eb-87e9-41f6-a408-478fb64d873c","resolution":{"observed_at":"2026-08-06T17:02:45.685674Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:45.417478Z","title":"Improving multi-turn tool use with reinforcement learning","venue":null,"work_id":"8a651294-d136-4d62-9af9-982dd4a8aa4b","year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":62,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:44.107767Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:9ff5b3352b951c14f02b9caac484bafa15a157144965bc78b029bf3663c12a63","observation_id":"d039b341-9518-4f29-852b-e212aa614854","resolution":{"observed_at":"2026-08-06T17:02:45.487366Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:45.189480Z","title":null,"venue":null,"work_id":"a1dd7165-38e7-460a-b0c2-a4cae3eb4d68","year":null},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":64,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:44.231115Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:b25e010563493ef2e57fbb74d3eb55416e53de58dc2b87df6da92395cf75c025","observation_id":"dbc77750-81a4-430c-88d1-3a78e80d79ed","resolution":{"observed_at":"2026-08-06T17:02:45.320141Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:45.009961Z","title":"We also ask it to generate sample tensor sizes for the task","venue":null,"work_id":"19b45797-c65b-419a-879b-feef8b2ef0b9","year":null},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":65,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:44.338778Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:042dd7cf87ae6cfd3ed663924887a5b5d083e38df5b36cd1513f64ef0bf5f06a","observation_id":"291afe17-dd9b-485d-886f-3edd0c74f5d8","resolution":{"observed_at":"2026-08-06T17:02:45.072245Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:44.865116Z","title":null,"venue":null,"work_id":"ffaa74e4-1aa3-4d91-91d2-c247d41083b5","year":null},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":66,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:44.408209Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:25d833294ffb3b0a9e1823095cee57c432898094855239d924cc45f87da7aa22","observation_id":"6da7b0f1-4772-4d92-9e7c-df10dc8e5c98","resolution":{"observed_at":"2026-08-06T17:02:44.927750Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:44.621774Z","title":"\"\" 7 Simple model that performs Layer Normalization. 8","venue":null,"work_id":"1554409f-0198-45a6-ac09-8cbbb9573a85","year":null},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":67,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:44.482544Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:5d5f45bdb51533075a195a7acb612db05a629c88d4b702c92a0cf7e737b853f2","observation_id":"06f89d91-28ad-416e-a5b0-bb39d977de76","resolution":{"observed_at":"2026-08-06T17:02:44.729040Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-15T06:32:42.880941+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T17:02:38.223669Z","title":null,"venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels","version":1},"reference_index":2025,"source":"pdf_text","source_observed_at":"2026-08-06T17:02:38.223669Z"},"links":{"citing_paper":"/paper/2507.11948"},"observation_digest":"sha256:6671f981a347aa4c12f7c65cf320cef5d58b3c8ce7451b2a7c5a00c195ab0d8c","observation_id":"6fea800c-415e-45cc-9ebe-8f3dec6e0287","resolution":{"observed_at":"2026-08-06T17:02:38.223669Z","resolver_source":null,"status":"parse_uncertain"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"paper":{"arxiv_id":"2507.11948","last_updated":"2025-07-16T06:33:07Z","latest_version":1,"primary_category":"cs.LG","snapshot_observed_at":"2026-08-09T05:05:38.345869Z","submitted_at":"2025-07-16T06:33:07Z","title":"Kevin: Multi-Turn RL for Generating CUDA Kernels"},"reference_resolution":{"displayed":67,"state_counts":{"malformed_identifier":0,"metadata_mismatch":0,"parse_uncertain":1,"unresolved":36,"verified_exact":0,"verified_fuzzy":30},"total_outbound_references":67},"refusal":"A citation records a reference. It does not transfer a finding from one paper to another.","schema":"pith.paper-citation-record.v1","standing_sources":[{"observed_at":"2026-08-15T06:32:42.880941+00:00","source":"crossref"},{"observed_at":"2026-08-15T06:32:39.529945+00:00","source":"retraction_watch"}],"thesis":"As of 16 August 2026, this Paper Citation Record lists 67 of 67 outbound references and 28 inbound Pith citation observations for arXiv:2507.11948."}