{"as_of":"2026-08-19T06:26:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:2ca52fd30c3ee0b0b5bc3bcf5725730aed9cd720721fefd098e304039d6d23bc","coverage":[{"denominator":50,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":50,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-06T15:53:17.648188Z","state":"measured"},{"denominator":84,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":84,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-18T06:34:40.430872+00:00","state":"measured"},{"denominator":34,"lane":"inbound_itemization","note":"Pith citing papers itemized under the disclosed page cap.","records_observed":34,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-14T04:13:34.537894Z","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":"2026-07-10T12:15:01.137692Z","state":"measured"}],"external_citation_measurements":[],"inbound":[{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-08-05T16:57:08.146810Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2508.17511","last_updated":"2025-08-24T20:23:08Z","snapshot_observed_at":"2026-08-14T01:23:22.506041Z","submitted_at":"2025-08-24T20:23:08Z","title":"School of Reward Hacks: Hacking harmless tasks generalizes to misaligned behavior in LLMs","version":1},"reference_index":10,"source":"arxiv_source","source_observed_at":"2026-08-05T16:57:08.146810Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2508.17511"},"observation_digest":"sha256:94ecf1688b8ca261589cb4f8cb2cd3114b1c3cce5e23c0ab73b5c18b664d96dc","observation_id":"78da8815-e0ee-4ce5-bc17-5d4387958ff1","resolution":{"observed_at":"2026-08-05T16:57:08.146810Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2601.03846","last_updated":"2026-04-19T20:12:38Z","snapshot_observed_at":"2026-08-14T23:46:24.848920Z","submitted_at":"2026-01-07T12:07:48Z","title":"When Numbers Start Talking: Implicit Numerical Coordination Among LLM-Based Agents","version":2},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-05-16T16:39:34.129361Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2601.03846"},"observation_digest":"sha256:0521593e7a8453445318d489975b184d39421c5daff8bf3e9bd20b3c2f682647","observation_id":"6fa4f8e9-0d44-43dc-bab7-8b10a0ffafd1","resolution":{"observed_at":"2026-05-16T16:41:06.209227Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2604.17614","last_updated":"2026-04-19T20:58:25Z","snapshot_observed_at":"2026-08-14T08:41:23.060090Z","submitted_at":"2026-04-19T20:58:25Z","title":"Characterizing Model-Native Skills","version":1},"reference_index":48,"source":"arxiv_source","source_observed_at":"2026-05-10T05:42:49.694715Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2604.17614"},"observation_digest":"sha256:ac2c7e560d04f52633465e0bb9999e251b9b689db46b22543d0da81274b05300","observation_id":"9757a501-16f4-48fb-a4ec-afa48f6522c4","resolution":{"observed_at":"2026-05-10T06:06:19.531920Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2604.17803","last_updated":"2026-04-20T04:51:39Z","snapshot_observed_at":"2026-08-17T04:22:21.600373Z","submitted_at":"2026-04-20T04:51:39Z","title":"Adversarial Arena: Crowdsourcing Data Generation through Interactive Competition","version":1},"reference_index":50,"source":"arxiv_source","source_observed_at":"2026-05-10T04:55:43.987116Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2604.17803"},"observation_digest":"sha256:85b18edeffd7e1eebec709154598062ed904ab1380103a5a27fc6bc86ba86be6","observation_id":"0ac5e9a4-e53d-4b9c-8af8-2cf7b7e9ec0e","resolution":{"observed_at":"2026-05-10T11:10:09.422596Z","resolver_source":"arxiv_id","status":"metadata_mismatch"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2604.23065","last_updated":"2026-07-01T23:54:28Z","snapshot_observed_at":"2026-08-15T11:09:06.968819Z","submitted_at":"2026-04-24T23:28:38Z","title":"What Should Frontier AI Developers Disclose About Internal Deployments?","version":1},"reference_index":19,"source":"arxiv_source","source_observed_at":"2026-05-08T09:33:47.869030Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2604.23065"},"observation_digest":"sha256:43e78387fb23f10046d8dc59fdc171add38cd6e75d0bc3e7ce48de76d08d29d7","observation_id":"40b1587c-8136-4695-a00b-40bbbc5371f1","resolution":{"observed_at":"2026-05-11T20:21:08.776819Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2604.25779","last_updated":"2026-04-28T15:46:18Z","snapshot_observed_at":"2026-08-13T04:03:15.102030Z","submitted_at":"2026-04-28T15:46:18Z","title":"Sustained Gradient Alignment Mediates Subliminal Learning in a Multi-Step Setting: Evidence from MNIST Auxiliary Logit Distillation Experiment","version":1},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-05-07T16:27:44.398710Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2604.25779"},"observation_digest":"sha256:2d22e209be29ccc84e857b62f74bf2479cb17248372850bae40e440d2ceda91d","observation_id":"d1cf6b6f-9f58-4f65-8e7d-edc671d9e18b","resolution":{"observed_at":"2026-05-11T23:41:18.235484Z","resolver_source":"arxiv_id","status":"metadata_mismatch"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2604.25783","last_updated":"2026-08-16T17:08:00Z","snapshot_observed_at":"2026-08-19T06:14:09.613186Z","submitted_at":"2026-04-28T15:51:55Z","title":"Subliminal Steering: Stronger Encoding of Hidden Signals","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-05-07T16:04:55.601088Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2604.25783"},"observation_digest":"sha256:e27e7363d24ab94b226e892b853c8197828c20e04528d4e9daa95eb8b6b770be","observation_id":"51e6e908-355f-477b-b37d-3d97c0dd6407","resolution":{"observed_at":"2026-05-11T23:56:12.065621Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2605.01130","last_updated":"2026-05-01T22:01:31Z","snapshot_observed_at":"2026-08-11T14:06:58.954849Z","submitted_at":"2026-05-01T22:01:31Z","title":"Iterative Finetuning is Mostly Idempotent","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-05-09T19:04:36.210200Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2605.01130"},"observation_digest":"sha256:b2b2db5ba1cdac5eed813272bb7a3e7a735b1c212606b8af1268cce84a6859ab","observation_id":"dc9c7883-3bca-4edb-9f00-540de9961269","resolution":{"observed_at":"2026-05-11T15:51:44.564144Z","resolver_source":"arxiv_id","status":"metadata_mismatch"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2605.02751","last_updated":"2026-05-10T16:16:37Z","snapshot_observed_at":"2026-08-13T04:03:17.764596Z","submitted_at":"2026-05-04T15:54:46Z","title":"Mitigating Misalignment Contagion by Steering with Implicit Traits","version":1},"reference_index":3,"source":"arxiv_source","source_observed_at":"2026-05-08T18:26:07.863552Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2605.02751"},"observation_digest":"sha256:393fa47f8e52467e13f47f30674866b7790c666bcfc0d3ddf89640a9fb2a6b8f","observation_id":"87088ca1-d07a-42b2-84ff-d1f53d53ad76","resolution":{"observed_at":"2026-05-09T06:25:48.481604Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2605.02751","last_updated":"2026-05-10T16:16:37Z","snapshot_observed_at":"2026-08-13T04:03:17.764596Z","submitted_at":"2026-05-04T15:54:46Z","title":"Mitigating Misalignment Contagion by Steering with Implicit Traits","version":2},"reference_index":3,"source":"arxiv_source","source_observed_at":"2026-05-12T02:18:07.412933Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2605.02751"},"observation_digest":"sha256:1221846ccc405c3a3f1e2a5c2993890ad175e9a435a4da5fb39c4f53544c1e4a","observation_id":"f40c78ed-220f-4f98-bf16-ffb4e0aa0c12","resolution":{"observed_at":"2026-05-12T07:41:44.277089Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2605.02930","last_updated":"2026-04-27T18:07:58Z","snapshot_observed_at":"2026-08-11T11:08:53.269135Z","submitted_at":"2026-04-27T18:07:58Z","title":"Analysis and Explainability of LLMs Via Evolutionary Methods","version":1},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-05-09T20:37:40.932811Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2605.02930"},"observation_digest":"sha256:1890da2415490260dcc81025d185e65f1be9d3824cbaa231785c167373cd14b0","observation_id":"17793f16-95b3-4c9d-b40d-894cfed7119f","resolution":{"observed_at":"2026-05-11T15:11:05.187751Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2605.06846","last_updated":"2026-06-02T16:52:04Z","snapshot_observed_at":"2026-08-16T04:00:04.477598Z","submitted_at":"2026-05-07T18:48:09Z","title":"Narrow Secret Loyalty Dodges Black-Box Audits","version":1},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-05-11T00:53:49.010929Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2605.06846"},"observation_digest":"sha256:80a7b77118c464b9c21cef40244a409fe943c6462736fbf528167997736badb8","observation_id":"8aa7bfb7-6d61-4737-a5da-1a61372704ee","resolution":{"observed_at":"2026-05-11T05:00:57.061982Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2605.06846","last_updated":"2026-06-02T16:52:04Z","snapshot_observed_at":"2026-08-16T04:00:04.477598Z","submitted_at":"2026-05-07T18:48:09Z","title":"Narrow Secret Loyalty Dodges Black-Box Audits","version":2},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-05-13T06:07:42.567241Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2605.06846"},"observation_digest":"sha256:53f85bf1437e68e3cd3266d26b66333c243a873e6fad26470774e04bfc287104","observation_id":"188e018b-198b-49b3-9c63-9724478256bf","resolution":{"observed_at":"2026-05-13T06:12:22.902036Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2605.06846","last_updated":"2026-06-02T16:52:04Z","snapshot_observed_at":"2026-08-16T04:00:04.477598Z","submitted_at":"2026-05-07T18:48:09Z","title":"Narrow Secret Loyalty Dodges Black-Box Audits","version":3},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-06-30T23:02:20.906168Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2605.06846"},"observation_digest":"sha256:1733e3cb0e7589be5f6c99460455ba615de7e405fc311c5a3920020728747492","observation_id":"a66c1554-7c7a-4240-838a-db3753baa41a","resolution":{"observed_at":"2026-06-30T23:05:07.332094Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2605.11290","last_updated":"2026-05-11T22:17:51Z","snapshot_observed_at":"2026-08-14T12:50:52.304597Z","submitted_at":"2026-05-11T22:17:51Z","title":"ReAD: Reinforcement-Guided Capability Distillation for Large Language Models","version":1},"reference_index":8,"source":"pdf_text","source_observed_at":"2026-05-13T01:45:26.316655Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2605.11290"},"observation_digest":"sha256:2b5f4313706a76ce25c6a3a415ce5a3290cfcbefc5778c7ef18bdf91c9fa2714","observation_id":"56938f5f-e6e1-499f-a0f9-49ad787c3a36","resolution":{"observed_at":"2026-05-13T01:47:04.068686Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2605.12199","last_updated":"2026-05-12T14:37:55Z","snapshot_observed_at":"2026-08-18T07:19:17.448191Z","submitted_at":"2026-05-12T14:37:55Z","title":"Overtrained, Not Misaligned","version":1},"reference_index":19,"source":"arxiv_source","source_observed_at":"2026-05-13T06:45:52.544674Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2605.12199"},"observation_digest":"sha256:4e76198026707df4dd5624e84fc373802e4918d39dba4d82a22d0e750329e060","observation_id":"f580fcd9-c1ca-470a-ba6f-3d598d257641","resolution":{"observed_at":"2026-05-13T06:47:26.314846Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2605.12798","last_updated":"2026-05-12T22:27:32Z","snapshot_observed_at":"2026-08-15T02:23:08.985643Z","submitted_at":"2026-05-12T22:27:32Z","title":"Emergent and Subliminal Misalignment Through the Lens of Data-Mediated Transfer","version":1},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-05-14T20:36:44.104197Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2605.12798"},"observation_digest":"sha256:0e057b2477b943ec6f65c445a2ccc3b9b3451ad4200cbe20a8abebfcc231de86","observation_id":"a6e71088-47eb-47e2-8f85-a53fca9f8bae","resolution":{"observed_at":"2026-05-14T20:37:58.509248Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.00831","last_updated":"2026-05-30T18:05:35Z","snapshot_observed_at":"2026-07-06T23:41:30.158879Z","submitted_at":"2026-05-30T18:05:35Z","title":"Subliminal Learning is a LoRA Artifact","version":1},"reference_index":1,"source":"arxiv_source","source_observed_at":"2026-06-28T18:28:42.513733Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.00831"},"observation_digest":"sha256:a4fcc94efb56098912cf7282e56c8906af9ca2ade7c5c084ca95838dc344a558","observation_id":"ffab0e9e-4852-4d02-837a-c523ac81c0c8","resolution":{"observed_at":"2026-06-28T20:42:37.592715Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.03810","last_updated":"2026-06-03T10:22:34Z","snapshot_observed_at":"2026-08-15T15:37:28.872815Z","submitted_at":"2026-06-02T15:54:24Z","title":"Consistency Training Can Entrench Misalignment","version":2},"reference_index":16,"source":"arxiv_source","source_observed_at":"2026-06-28T10:07:31.153337Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.03810"},"observation_digest":"sha256:fe750e9ddc30b53029b4d9bdbad826437ac79c784809be1d66f54e20d3a08ee9","observation_id":"8dcf6e04-e44e-4225-8c55-609add21e171","resolution":{"observed_at":"2026-07-02T03:26:28.603516Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.06315","last_updated":"2026-06-04T15:54:34Z","snapshot_observed_at":"2026-08-13T04:02:14.220214Z","submitted_at":"2026-06-04T15:54:34Z","title":"LLM Self-Recognition: Steering and Retrieving Activation Signatures","version":1},"reference_index":2,"source":"arxiv_source","source_observed_at":"2026-06-28T01:41:03.518190Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.06315"},"observation_digest":"sha256:5870e2fae5d67493db64b91e2ecc7fef7f5c60cf68dab8b97a5bc7685ad49f1f","observation_id":"98c3d4b0-50df-43e1-b7b1-11a1caa39171","resolution":{"observed_at":"2026-07-02T12:56:57.433832Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.06667","last_updated":"2026-07-06T17:01:35Z","snapshot_observed_at":"2026-08-12T12:19:27.839519Z","submitted_at":"2026-06-04T19:32:00Z","title":"The Piggyback Hypothesis of Generalization: Explaining and Mitigating Emergent Misalignment","version":1},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-06-28T01:34:23.382705Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.06667"},"observation_digest":"sha256:eaa484005cb91a350f37443a7204cd898ea1f1b97d30b7b0ffb46c14f370424e","observation_id":"5b28e095-6a9b-476b-9235-cb6053f576ac","resolution":{"observed_at":"2026-07-02T13:06:59.146610Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-12T14:59:04.152474Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2606.06667","last_updated":"2026-07-06T17:01:35Z","snapshot_observed_at":"2026-08-12T12:19:27.839519Z","submitted_at":"2026-06-04T19:32:00Z","title":"The Piggyback Hypothesis of Generalization: Explaining and Mitigating Emergent Misalignment","version":2},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-07-12T14:59:04.152474Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.06667"},"observation_digest":"sha256:d2e0bf46b6752f9e3411c14e2ce004c8e0c5ec850d89a740f491fc4fef876e85","observation_id":"90c699d8-1b52-408e-b152-522de8ea17a8","resolution":{"observed_at":"2026-07-12T14:59:04.152474Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.07602","last_updated":"2026-05-29T09:31:25Z","snapshot_observed_at":"2026-07-06T23:47:15.041928Z","submitted_at":"2026-05-29T09:31:25Z","title":"Sample-Efficient Post-Training for LEGO Spatial-Physics Reasoning","version":1},"reference_index":65,"source":"arxiv_source","source_observed_at":"2026-06-28T23:38:38.127345Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.07602"},"observation_digest":"sha256:8d381db8b3d393b418a85d6411867faec10e1400631ec2cc636bc9caa9ba5482","observation_id":"3dea853d-1ba4-432b-aabd-33ff234471a1","resolution":{"observed_at":"2026-06-28T23:42:49.502198Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.11270","last_updated":"2026-06-26T21:36:00Z","snapshot_observed_at":"2026-08-13T12:59:15.970811Z","submitted_at":"2026-06-09T06:52:49Z","title":"Quantifying Subliminal Behavioral Transfer Ratios in Language Model Distillation","version":1},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-06-27T13:45:13.320426Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.11270"},"observation_digest":"sha256:2d4fae752ac1006059db4408ebc499f20bbde9faa9d8dbd905ed83e004c984e3","observation_id":"bc748230-e09d-4318-9bc3-3bbde11e7e04","resolution":{"observed_at":"2026-07-03T04:37:37.317342Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.11270","last_updated":"2026-06-26T21:36:00Z","snapshot_observed_at":"2026-08-13T12:59:15.970811Z","submitted_at":"2026-06-09T06:52:49Z","title":"Quantifying Subliminal Behavioral Transfer Ratios in Language Model Distillation","version":2},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-06-30T11:15:21.946819Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.11270"},"observation_digest":"sha256:87c80254271917e3b3f9a6d61391d797949da4972a73d246bebe569c930e78cd","observation_id":"c7849f53-caad-47c5-a841-a73812005f57","resolution":{"observed_at":"2026-06-30T11:24:38.477330Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.12360","last_updated":"2026-06-10T17:31:16Z","snapshot_observed_at":"2026-08-15T04:26:22.769006Z","submitted_at":"2026-06-10T17:31:16Z","title":"Anatomy of Post-Training: Using Interpretability to Characterize Data and Shape the Learning Signal","version":1},"reference_index":25,"source":"arxiv_source","source_observed_at":"2026-06-27T10:32:57.295159Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.12360"},"observation_digest":"sha256:a876c218c089a90dbc9fd6a521bef529d09e6d34589dad481e8f4345c5c6fa8d","observation_id":"4a1fe570-b980-444e-aeb7-1dcea8de1d82","resolution":{"observed_at":"2026-07-03T09:07:47.865564Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.12385","last_updated":"2026-06-10T17:47:59Z","snapshot_observed_at":"2026-08-15T14:37:28.125966Z","submitted_at":"2026-06-10T17:47:59Z","title":"Which Models Are Our Models Built On? Auditing Invisible Dependencies in Modern LLMs","version":1},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-06-27T09:57:14.328157Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.12385"},"observation_digest":"sha256:52439ffe1b26c2189a7f8053362b7e6a7260b7ac74e18834712e305d8f98b8ad","observation_id":"e72b5fd4-22cc-4ebe-ac0b-331a1ec72b46","resolution":{"observed_at":"2026-07-03T10:37:56.589387Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.20512","last_updated":"2026-06-19T10:59:07Z","snapshot_observed_at":"2026-08-15T13:55:22.298822Z","submitted_at":"2026-06-18T17:30:15Z","title":"Probe-and-Refine Tuning of Repository Guidance for Coding Agents","version":2},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-06-26T15:58:40.881243Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.20512"},"observation_digest":"sha256:1dfd02aae1e966ecf2350568d7914474c6b822e5d4c231bdeb55e7d93b00d9a8","observation_id":"63f1c808-701f-4288-8d42-b8a70539c942","resolution":{"observed_at":"2026-07-04T05:29:35.754233Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2606.22019","last_updated":"2026-06-20T12:48:31Z","snapshot_observed_at":"2026-08-06T19:21:18.734073Z","submitted_at":"2026-06-20T12:48:31Z","title":"Channel Location Constrains the Auditability of Subliminal Learning","version":1},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-06-26T11:52:03.948568Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2606.22019"},"observation_digest":"sha256:cc99704acf183fdd2a4db9ef884f46676d625455bbce81906ddca7625af587c3","observation_id":"88be5b9d-b074-475b-99f5-f940d587f5a2","resolution":{"observed_at":"2026-07-04T08:19:44.436777Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":"2507.14805","doi":"10.48550/arxiv.2507.14805","metadata_source":"arxiv_reference","pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-10T12:15:01.137692Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data.arXiv preprint arXiv:2507.14805","venue":null,"work_id":"5fc1e083-accb-4ae7-bf99-4a5e72ca356f","year":2025},"citing_paper":{"arxiv_id":"2607.01033","last_updated":"2026-07-01T15:01:30Z","snapshot_observed_at":"2026-08-17T12:21:30.498323Z","submitted_at":"2026-07-01T15:01:30Z","title":"The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology","version":1},"reference_index":11,"source":"arxiv_source","source_observed_at":"2026-07-02T15:57:48.589980Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2607.01033"},"observation_digest":"sha256:e3a5c8a4923c51a930b1079d49afce0ef1600f8f9fc73b936538bfda4125cda4","observation_id":"eb938279-88f7-4753-a87e-8ff6ecc6d4d6","resolution":{"observed_at":"2026-07-02T16:07:08.140193Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-07-11T20:51:42.052454Z","title":"arXiv preprint arXiv:2507.14805 , year=","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2607.04222","last_updated":"2026-07-05T10:28:04Z","snapshot_observed_at":"2026-08-19T04:11:28.929417Z","submitted_at":"2026-07-05T10:28:04Z","title":"Unsupervised Features Mining via Activation Geometry","version":1},"reference_index":1,"source":"arxiv_source","source_observed_at":"2026-07-11T20:51:42.052454Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2607.04222"},"observation_digest":"sha256:6c14d09cb0c9f8d2d1b2334136e55f0df2bfb2c4a7b6b2eda8f76fa59ec25bd2","observation_id":"6c35d487-4c19-40ce-b04f-fc67b8a7cce2","resolution":{"observed_at":"2026-07-11T20:51:42.052454Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-08-01T07:46:09.323606Z","title":"arXiv preprint arXiv:2507.14805 , year =","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2607.21356","last_updated":"2026-07-23T14:19:28Z","snapshot_observed_at":"2026-08-14T00:44:11.658093Z","submitted_at":"2026-07-23T14:19:28Z","title":"Emergent Misalignment Recruits a Pre-existing Persona Subspace","version":1},"reference_index":79,"source":"arxiv_source","source_observed_at":"2026-08-01T07:46:09.323606Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2607.21356"},"observation_digest":"sha256:0052a7f092860cf8aa11dcf87ef772d1754b47ccea45599e3fde231358aa5ee4","observation_id":"7fa3c2e8-c11c-4269-8911-963ae569502c","resolution":{"observed_at":"2026-08-01T07:46:09.323606Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-08-01T00:38:41.279652Z","title":"Subliminal learning: Language models transmit behavioral traits via hidden signals in data","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2607.26173","last_updated":"2026-07-28T18:29:45Z","snapshot_observed_at":"2026-08-12T12:20:37.312304Z","submitted_at":"2026-07-28T18:29:45Z","title":"Shared SFT Lessons Across Alignment, Model Organisms, and Toy Models","version":1},"reference_index":8,"source":"pdf_text","source_observed_at":"2026-08-01T00:38:41.279652Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2607.26173"},"observation_digest":"sha256:e0b88f9f22d97ac5c437a32063435af38bf6d04886d223eeed7f9ef312311c73","observation_id":"b5dec608-6606-45fc-abbd-a41cec86ab9a","resolution":{"observed_at":"2026-08-01T00:38:41.279652Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2507.14805","snapshot_observed_at":"2026-08-14T04:13:34.537894Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data, July 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2608.10218","last_updated":"2026-08-10T20:37:57Z","snapshot_observed_at":"2026-08-18T18:59:21.085252Z","submitted_at":"2026-08-10T20:37:57Z","title":"Mind Viruses: Self-Propagating Ideas in Multi-Agent LLM Systems","version":1},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-14T04:13:34.537894Z"},"links":{"cited_paper":"/paper/2507.14805","citing_paper":"/paper/2608.10218"},"observation_digest":"sha256:040205e20a58ee18ac3ec7550b6bcbcd9618bfe979138fc12688f77df7e758e1","observation_id":"2c681241-f1ae-4436-add0-4fa973d9c1e6","resolution":{"observed_at":"2026-08-14T04:13:34.537894Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"links":{"evidence":"/evidence","html":"/paper/2507.14805/citation-record","integrity":"/paper/2507.14805/integrity","json":"/paper/2507.14805/citation-record.json","paper":"/paper/2507.14805"},"outbound":[{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T15:53:16.207156Z","title":"write newline","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":1,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:16.207156Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:727b582815ac847b2d50957ef4bdb5079cb7316f82d57e1dd573695d3409e5be","observation_id":"f63e51f4-4ac8-4f8f-af7e-f7825764911d","resolution":{"observed_at":"2026-08-06T15:53:16.207156Z","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":"2026-08-06T15:53:18.407389Z","title":"Claude 3.7 Sonnet and Claude Code","venue":null,"work_id":"4c1c9720-8cd2-4ebb-923a-507dd90ef886","year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":2,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:16.250854Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:6f55334095bcba63e91dfefd0528fd9a8e2a07f54abfc79a09ec9c804c912107","observation_id":"bf2b658f-3e09-46ab-90f5-debcd43fc07a","resolution":{"observed_at":"2026-08-06T15:53:18.410390Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2503.11926","last_updated":"2025-03-14T23:50:34Z","snapshot_observed_at":"2026-08-17T15:42:03.133713Z","submitted_at":"2025-03-14T23:50:34Z","title":"Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2503.11926","snapshot_observed_at":"2026-08-06T15:53:16.382494Z","title":"Monitoring reasoning models for misbehavior and the risks of promoting obfuscation","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":3,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:16.382494Z"},"links":{"cited_paper":"/paper/2503.11926","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:2713d3aeb74184ac8ee0be5149090586d25a6df4cebefc4b4a8727a650a0c50e","observation_id":"8840d800-0164-45bf-81ff-88cfcd9c08af","resolution":{"observed_at":"2026-08-06T15:53:16.382494Z","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":"2026-08-06T15:53:18.398391Z","title":"Hiding images in plain sight: Deep steganography","venue":null,"work_id":"f50f97aa-2892-4e1e-ba76-42e7468db7dc","year":2017},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":4,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:16.488421Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:a12c6dcdc3f416585ba80d59d61292a5e4f1b3269aa3bc6112043afd210b0c16","observation_id":"cc014cdd-2f1b-4d55-b8f3-9897003da5dd","resolution":{"observed_at":"2026-08-06T15:53:18.401781Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:16.587780Z","title":"Emergent misalignment: Narrow finetuning can produce broadly misaligned llms","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":5,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:16.587780Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:628833fd7aa29d90141e34c6f0a84ed2c0b03d985bf0588fbc9af792f2a2f237","observation_id":"c51ce286-65df-4d22-aabd-e1b6c55b06cc","resolution":{"observed_at":"2026-08-06T15:53:16.587780Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1206.6389","last_updated":"2013-03-25T10:16:36Z","snapshot_observed_at":"2026-08-15T00:55:08.947472Z","submitted_at":"2012-06-27T19:59:59Z","title":"Poisoning Attacks against Support Vector Machines","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1206.6389","snapshot_observed_at":"2026-08-06T15:53:16.678855Z","title":"Poisoning attacks against support vector machines","venue":null,"work_id":null,"year":2012},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":6,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:16.678855Z"},"links":{"cited_paper":"/paper/1206.6389","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:68eaae985af7216a223c414ae6f4e7ed467b05639dbbb560aff1fbcc38753a85","observation_id":"1d532aa2-e1bd-4d9c-8f0e-08b92d63135f","resolution":{"observed_at":"2026-08-06T15:53:16.678855Z","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":"2026-08-06T15:53:16.735243Z","title":"Language models are few-shot learners","venue":null,"work_id":null,"year":1901},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":7,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:16.735243Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:02035c92575859141bf92a3923705c5ba8e9cc96e7ac11408075543917f92663","observation_id":"1b4bb6c4-cf30-47cb-994c-b0d48d18effe","resolution":{"observed_at":"2026-08-06T15:53:16.735243Z","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":"2026-08-06T15:53:18.383686Z","title":"An information-theoretic model for steganography","venue":null,"work_id":"474bb2c6-02be-484c-8634-0657e22ea98b","year":1998},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":8,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:16.876524Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:e1b78cdff6347e60e5b443c77f5ea734451d9146cc97683be8c1500541326420","observation_id":"5141350f-19a9-46e2-9bc5-3dfb84cb14ce","resolution":{"observed_at":"2026-08-06T15:53:18.387208Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:16.927782Z","title":"Undetectable watermarks for language models","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":9,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:16.927782Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:2cd79047933a5c048aef145b87fd2b305ffc70714b2a4fb8eb56582be545f515","observation_id":"a0725437-470a-474b-83c9-0a95df38cb18","resolution":{"observed_at":"2026-08-06T15:53:16.927782Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2506.13206","last_updated":"2025-07-10T08:27:27Z","snapshot_observed_at":"2026-08-16T05:53:38.303108Z","submitted_at":"2025-06-16T08:10:04Z","title":"Thought Crime: Backdoors and Emergent Misalignment in Reasoning Models","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2506.13206","snapshot_observed_at":"2026-08-06T15:53:17.063185Z","title":"Thought crime: Backdoors and emergent misalignment in reasoning models","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":10,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.063185Z"},"links":{"cited_paper":"/paper/2506.13206","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:85e021cddef365e5cea1e73d0873f1c166ca2832011baabe9f07a81f22cc99a3","observation_id":"ee0e01b8-c67a-43fc-ab24-a981e4eea63c","resolution":{"observed_at":"2026-08-06T15:53:17.063185Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2110.14168","last_updated":"2021-11-18T00:23:45Z","snapshot_observed_at":"2026-08-14T02:43:01.480086Z","submitted_at":"2021-10-27T04:49:45Z","title":"Training Verifiers to Solve Math Word Problems","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2110.14168","snapshot_observed_at":"2026-08-06T15:53:17.135720Z","title":"Training verifiers to solve math word problems","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":11,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.135720Z"},"links":{"cited_paper":"/paper/2110.14168","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:b2c0feee2d68b04c114ad534103348989be2fba05554cddf1ecee40e6e52b0a4","observation_id":"e72a0396-7f77-4a2b-930b-31ffb1840769","resolution":{"observed_at":"2026-08-06T15:53:17.135720Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2406.10162","last_updated":"2024-06-29T00:28:47Z","snapshot_observed_at":"2026-08-18T14:49:13.384384Z","submitted_at":"2024-06-14T16:26:20Z","title":"Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2406.10162","snapshot_observed_at":"2026-08-06T15:53:17.228125Z","title":"Sycophancy to subterfuge: Investigating reward-tampering in large language models","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":12,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.228125Z"},"links":{"cited_paper":"/paper/2406.10162","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:46add4244411c13d9648d8c62e1119204eac5f5a4f42029443947c47c3d560db","observation_id":"629a9485-7544-4cee-b5ef-0c57c87e5840","resolution":{"observed_at":"2026-08-06T15:53:17.228125Z","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":"2026-08-06T15:53:17.301215Z","title":"RAFT : Reward ranked finetuning for generative foundation model alignment","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":13,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.301215Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:069c182d4fff7806050f3c2b69d3ee0a6f2838fe932d311ca704676be26cf403","observation_id":"2eed0a5b-05cf-49c8-bb2d-01acc31a89f8","resolution":{"observed_at":"2026-08-06T15:53:17.301215Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2503.01926","last_updated":"2025-06-03T08:19:31Z","snapshot_observed_at":"2026-08-16T17:24:15.313343Z","submitted_at":"2025-03-02T12:10:17Z","title":"Unnatural Languages Are Not Bugs but Features for LLMs","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2503.01926","snapshot_observed_at":"2026-08-06T15:53:17.423707Z","title":"Unnatural languages are not bugs but features for llms","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":14,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.423707Z"},"links":{"cited_paper":"/paper/2503.01926","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:c28ab9c9aa6fb3e0d68fc19c2e3972dcd51140abc7264fae3e9c8403fd875256","observation_id":"7a3946e6-fc79-4474-9153-926e494b5f9d","resolution":{"observed_at":"2026-08-06T15:53:17.423707Z","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":"2026-08-06T15:53:17.500345Z","title":"Born again neural networks","venue":null,"work_id":null,"year":2018},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":15,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.500345Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:99cdbfdd9e85a635de7b680ef25e238f20cd0c24536dc0fc7afc339ec8fdc834","observation_id":"16418560-22b8-47a5-adf5-35e20d2c56d5","resolution":{"observed_at":"2026-08-06T15:53:17.500345Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2412.14093","last_updated":"2024-12-20T02:22:19Z","snapshot_observed_at":"2026-08-13T11:12:58.507759Z","submitted_at":"2024-12-18T17:41:24Z","title":"Alignment faking in large language models","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2412.14093","snapshot_observed_at":"2026-08-06T15:53:17.506921Z","title":"Alignment faking in large language models","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":16,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.506921Z"},"links":{"cited_paper":"/paper/2412.14093","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:ac64a511ba565daeb123d9e714f2f4137b3403c6f3deccffac2e5e9770046600","observation_id":"08276008-d7f9-4aac-b4c5-93b8bd4d295f","resolution":{"observed_at":"2026-08-06T15:53:17.506921Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2412.16339","last_updated":"2025-01-08T20:11:59Z","snapshot_observed_at":"2026-08-17T07:21:20.458339Z","submitted_at":"2024-12-20T21:00:11Z","title":"Deliberative Alignment: Reasoning Enables Safer Language Models","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2412.16339","snapshot_observed_at":"2026-08-06T15:53:17.551777Z","title":"Deliberative alignment: Reasoning enables safer language models","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":17,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.551777Z"},"links":{"cited_paper":"/paper/2412.16339","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:4e2583b8e03f361643dcf13665702732d8aba1daa4fd3769fc93293498691076","observation_id":"efeff014-0da3-4387-89e2-6fc3a28658e7","resolution":{"observed_at":"2026-08-06T15:53:17.551777Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2501.12948","last_updated":"2026-01-04T03:57:36Z","snapshot_observed_at":"2026-08-15T12:33:55.451951Z","submitted_at":"2025-01-22T15:19:35Z","title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2501.12948","snapshot_observed_at":"2026-08-06T15:53:17.555074Z","title":"Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":18,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.555074Z"},"links":{"cited_paper":"/paper/2501.12948","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:f90ed51997eabe3fdd4299176815b29cba0df372fa2c286aff3b05cf2d0d2f0f","observation_id":"37e6a0df-3e08-44f6-8843-9b5cf7fc53b9","resolution":{"observed_at":"2026-08-06T15:53:17.555074Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1503.02531","last_updated":"2015-03-09T15:44:49Z","snapshot_observed_at":"2026-08-16T18:00:58.008096Z","submitted_at":"2015-03-09T15:44:49Z","title":"Distilling the Knowledge in a Neural Network","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1503.02531","snapshot_observed_at":"2026-08-06T15:53:17.558107Z","title":"Distilling the knowledge in a neural network","venue":null,"work_id":null,"year":2015},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":19,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.558107Z"},"links":{"cited_paper":"/paper/1503.02531","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:231a2264a6e13acdad9b3551e3479fc0f120ae6a42adecf5662ae10b64e10b9e","observation_id":"fae85c43-3d8d-486c-b5bc-0a85d8b4f292","resolution":{"observed_at":"2026-08-06T15:53:17.558107Z","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":"2026-08-06T15:53:17.561145Z","title":"Large language models are reasoning teachers","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":20,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.561145Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:dbe8222f48ab5c4120373d527348156109545375fd0c8d39c009c197e33a0891","observation_id":"14a26ee7-3d02-4a43-a6e8-8222df03ccf8","resolution":{"observed_at":"2026-08-06T15:53:17.561145Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2401.05566","last_updated":"2024-01-17T20:26:01Z","snapshot_observed_at":"2026-08-15T14:10:05.296241Z","submitted_at":"2024-01-10T22:14:35Z","title":"Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2401.05566","snapshot_observed_at":"2026-08-06T15:53:17.563959Z","title":"Sleeper agents: Training deceptive llms that persist through safety training","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":21,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.563959Z"},"links":{"cited_paper":"/paper/2401.05566","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:018782d0223d8db645d07de75e220d6e76022ec7cb8223a775031e37a15c2339","observation_id":"2cbee001-da72-4d4e-a773-0d9f381e8f7c","resolution":{"observed_at":"2026-08-06T15:53:17.563959Z","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":"2026-08-06T15:53:17.566996Z","title":"Adversarial examples are not bugs, they are features","venue":null,"work_id":null,"year":2019},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":22,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.566996Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:0f619f0de4ef22f887226fee71830818e4703702f490925fa3f9aa071485facb","observation_id":"c17fd8c2-2053-4ac0-95a2-25cbf4428744","resolution":{"observed_at":"2026-08-06T15:53:17.566996Z","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":"2026-08-06T15:53:18.352532Z","title":"Batch steganography and pooled steganalysis","venue":null,"work_id":"0d69699f-26c9-4909-b9dd-d326a9ec40d6","year":2006},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":23,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.569717Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:34cba60905904ab89d0e97dc6e9fe7c337ab36b59fcabb096deee71e6d7ef3b0","observation_id":"92bfb847-5174-434d-8bcb-732bf9279a39","resolution":{"observed_at":"2026-08-06T15:53:18.355496Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:18.343167Z","title":"A watermark for large language models","venue":null,"work_id":"8c22477c-5f4e-4a08-be5d-5f52341d93ed","year":2023},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":24,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.572529Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:7577fc271f4ea8fc8b239b60584bf3e12251f93c33ca01871fa37800007df311","observation_id":"1f3d71de-870f-4596-b922-ff28c821d1a8","resolution":{"observed_at":"2026-08-06T15:53:18.346285Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:17.575231Z","title":"Lecun, L","venue":null,"work_id":null,"year":1998},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":25,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.575231Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:73f875b93079c6811f2b3119b532f402cb47a3ee363b0a4fd414b081d8bc08a6","observation_id":"ec747bb5-7bc4-4a47-99b4-695a5a4091bd","resolution":{"observed_at":"2026-08-06T15:53:17.575231Z","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":"2506.06278","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-06T15:53:18.023407Z","title":"Distillation robustifies unlearning","venue":null,"work_id":"b34e09c9-ca91-40dc-a7f5-e05edd8396c9","year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":26,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.577940Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:ae031c51921f4563aa89e92cf5c3de118d622838cceec5f25d95303731aa02d6","observation_id":"fdc2360f-6b45-4ef8-bbe6-0d6bbda863dc","resolution":{"observed_at":"2026-08-06T15:53:18.028958Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:17.580834Z","title":"T ruthful QA : Measuring how models mimic human falsehoods","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":27,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.580834Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:5a11755f7c7bc9cff7399ef7114e72812b7057cde9e04cc9b1339a17733107f9","observation_id":"2a429e7c-2484-4951-a00e-b0b7beff4509","resolution":{"observed_at":"2026-08-06T15:53:17.580834Z","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":"2026-08-06T15:53:18.333557Z","title":"Secret collusion among ai agents: Multi-agent deception via steganography","venue":null,"work_id":"a5f0b5a7-0d33-4a67-a24b-826fd6ca1918","year":2024},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":28,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.584169Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:79ae10324c2d3f6245a2978f8f94ad241fc613331b50cafcb4b6e65444fb047c","observation_id":"5ab9744a-127a-4188-ab5b-f9c1781048a9","resolution":{"observed_at":"2026-08-06T15:53:18.337011Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:17.587017Z","title":"Self-imitation learning","venue":null,"work_id":null,"year":2018},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":29,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.587017Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:5c16b434d0488d070199c49e593b8fec1b24bb01aa70ac1c4900a645fca65286","observation_id":"20398895-6b0d-4020-8b98-b1212e4b4ac2","resolution":{"observed_at":"2026-08-06T15:53:17.587017Z","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":"2026-08-06T15:53:18.318760Z","title":"Hello gpt-4o","venue":null,"work_id":"e0252e11-e7c4-4a53-9752-1c4bf04070a4","year":2024},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":30,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.589776Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:acb25003f9b67421d9fcb9d53c92d6e8573056c51b495b3fa368ada067138776","observation_id":"db0ce348-8cbd-4a5c-af04-9f557e411058","resolution":{"observed_at":"2026-08-06T15:53:18.321433Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:18.308835Z","title":"Introducing gpt‑4.1 in the api","venue":null,"work_id":"4eddd843-354b-4079-9b85-a9fe8623d45c","year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":31,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.592548Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:73f24b9efefb67c6c86b3477f5a7a74b39b666cb3e69d484284dfdbd972d7de8","observation_id":"16bfdca0-cbd1-4f75-9993-3bc6af9eca83","resolution":{"observed_at":"2026-08-06T15:53:18.312345Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:18.299293Z","title":"Supervised Fine‑Tuning","venue":null,"work_id":"05b54b9a-9963-4189-8a91-bf53c39a34ce","year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":32,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.595293Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:286b0451fdc02f2476a2146c5b2719fc8917ff8ea523693134f4a66f657c956a","observation_id":"dc7c7e9b-249e-4063-a910-09ea43b87027","resolution":{"observed_at":"2026-08-06T15:53:18.302301Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:18.290355Z","title":null,"venue":null,"work_id":"d27ae156-7d0a-444e-91d5-ad94e6ddfb52","year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":33,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.598764Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:ef47bdc93f56987f811f2a08bbe94975036c00ce5f723fa78fa716e3dfb1c22c","observation_id":"24ba2781-a354-4efa-96cd-31e4611014dc","resolution":{"observed_at":"2026-08-06T15:53:18.293245Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:18.281569Z","title":"Model compression via distillation and quantization","venue":null,"work_id":"4c461cda-94a7-4c07-810e-6dbdb01350bb","year":2018},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":34,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.602079Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:7e3186a698bb253fff0d49bd0658a5dbed29f5ad82c43d8fb9eab73e8857ce3b","observation_id":"f0d4d640-9b47-47e4-a9fe-b78520c73eea","resolution":{"observed_at":"2026-08-06T15:53:18.284655Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:18.272011Z","title":"Hidden trigger backdoor attacks","venue":null,"work_id":"6078aab2-0b2d-4c49-be15-ecb75d15f308","year":2020},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":35,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.605729Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:886c5068bed4ec0532cf96c24b09fd15fd314bcb5597dec8e255482a4906b29f","observation_id":"e1567d1c-eea7-4043-bda7-c616f9517264","resolution":{"observed_at":"2026-08-06T15:53:18.274980Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:18.262465Z","title":"Poison frogs! targeted clean-label poisoning attacks on neural networks","venue":null,"work_id":"361b01b3-706b-4978-b93d-136f18d75d72","year":2018},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":36,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.608295Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:68443d342a9f8263cd48b775582be1c3b6e154a579c361266ef1980ed77ac84b","observation_id":"4109380c-1456-40ff-a8f9-0eef7c4665e2","resolution":{"observed_at":"2026-08-06T15:53:18.265794Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:17.610816Z","title":"Defining and characterizing reward gaming","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":37,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.610816Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:c4bf5d976616a8eda05408367f5562c2eaed193bd40c9d7adcb96f75265805e9","observation_id":"3eadc88d-b018-47aa-814e-b71724057a20","resolution":{"observed_at":"2026-08-06T15:53:17.610816Z","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":"2026-08-06T15:53:18.247526Z","title":"Certified defenses for data poisoning attacks","venue":null,"work_id":"9a4eb12e-7a32-4edc-a3c3-34af43906021","year":2017},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":38,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.613755Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:fe700b32e0b3e5c6b4dd2152f6c1ccc4f7c65c3d3f092d16c29e29e6fef9034b","observation_id":"e32f490d-bbda-4d00-9b1f-008045309321","resolution":{"observed_at":"2026-08-06T15:53:18.250319Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2506.11613","last_updated":"2025-06-13T09:34:25Z","snapshot_observed_at":"2026-08-17T09:29:38.464156Z","submitted_at":"2025-06-13T09:34:25Z","title":"Model Organisms for Emergent Misalignment","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2506.11613","snapshot_observed_at":"2026-08-06T15:53:17.616407Z","title":"Model organisms for emergent misalignment","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":39,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.616407Z"},"links":{"cited_paper":"/paper/2506.11613","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:dbfbbfafb680db7c108c8e07513833dc53494a7ceb8ff7d1d520dad8ef58388e","observation_id":"a98ca396-9c82-4303-b1e0-97e4e1cd2712","resolution":{"observed_at":"2026-08-06T15:53:17.616407Z","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":"2026-08-06T15:53:18.239085Z","title":"Concealed data poisoning attacks on nlp models","venue":null,"work_id":"47d7eba8-5592-4871-933b-252ec59fd798","year":2021},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":40,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.619301Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:7e7034271ce86a521fa2ce5c1b11f3c6a31271eefef78a2c1dd3a7fc5192e4d1","observation_id":"fef340f8-d08b-45bb-b8a9-879b06f1c9d0","resolution":{"observed_at":"2026-08-06T15:53:18.241890Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+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":"2026-08-06T15:53:17.622147Z","title":"Persona features control emergent misalignment","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":41,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.622147Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:24e926528268f3e48be0da6145bce5baf9fda7e174eac9f134969a6a61d959da","observation_id":"0deed0c4-01fb-4a7d-ad70-d01cc7889dce","resolution":{"observed_at":"2026-08-06T15:53:17.622147Z","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":"2026-08-06T15:53:17.624991Z","title":"Smith, Daniel Khashabi, and Hannaneh Hajishirzi","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":42,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.624991Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:45f87edcead5de85797e213d204e7b3a202efb87c2eeee3f2111efefda67f973","observation_id":"57cc65e3-e2bc-4dea-b8c6-776cd7fdddd4","resolution":{"observed_at":"2026-08-06T15:53:17.624991Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2406.01574","last_updated":"2024-11-06T02:54:00Z","snapshot_observed_at":"2026-08-17T13:17:07.963143Z","submitted_at":"2024-06-03T17:53:00Z","title":"MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark","version":6},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2406.01574","snapshot_observed_at":"2026-08-06T15:53:17.628302Z","title":"Mmlu-pro: A more robust and challenging multi-task language understanding benchmark, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":43,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.628302Z"},"links":{"cited_paper":"/paper/2406.01574","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:ff595b10880ca54a215162f19247976c0418c1ae3ea65720859c572c6d206065","observation_id":"30a969c8-e2ab-4366-beb1-6645b26abb7b","resolution":{"observed_at":"2026-08-06T15:53:17.628302Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2412.15115","last_updated":"2025-01-03T02:18:21Z","snapshot_observed_at":"2026-08-17T18:50:07.059564Z","submitted_at":"2024-12-19T17:56:09Z","title":"Qwen2.5 Technical Report","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2412.15115","snapshot_observed_at":"2026-08-06T15:53:17.631150Z","title":"Qwen2.5 technical report, 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":44,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.631150Z"},"links":{"cited_paper":"/paper/2412.15115","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:afec3ddc3722bd8099be9e37dbd070f9a97aae6933ceddc7d5a2064c2e25242c","observation_id":"ced52b3f-933e-4f50-9dab-971aa931c48e","resolution":{"observed_at":"2026-08-06T15:53:17.631150Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2305.17359","last_updated":"2023-10-04T16:36:09Z","snapshot_observed_at":"2026-08-16T15:29:06.598244Z","submitted_at":"2023-05-27T03:58:29Z","title":"DNA-GPT: Divergent N-Gram Analysis for Training-Free Detection of GPT-Generated Text","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2305.17359","snapshot_observed_at":"2026-08-06T15:53:17.633747Z","title":"Dna-gpt: Divergent n-gram analysis for training-free detection of gpt-generated text","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":45,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.633747Z"},"links":{"cited_paper":"/paper/2305.17359","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:c2d5145382862a8b010ab62629443093577894d1494144b47249850b676da51c","observation_id":"c79ad620-890e-4284-bf4a-edf970420b82","resolution":{"observed_at":"2026-08-06T15:53:17.633747Z","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":"2026-08-06T15:53:18.228896Z","title":"Protecting language generation models via invisible watermarking","venue":null,"work_id":"0849d949-fdd9-4b7b-ae47-f2e2d0171605","year":2023},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":46,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.636502Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:40c85b98f52371e279ab66ca851f89f99ece3423dde385387c6e2c74bb52579d","observation_id":"28a49ad1-5e1b-47f3-bc96-30d29f61940b","resolution":{"observed_at":"2026-08-06T15:53:18.233145Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"1909.01496","last_updated":"2019-09-03T23:15:19Z","snapshot_observed_at":"2026-08-18T08:22:49.298039Z","submitted_at":"2019-09-03T23:15:19Z","title":"Neural Linguistic Steganography","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1909.01496","snapshot_observed_at":"2026-08-06T15:53:17.639213Z","title":"Neural linguistic steganography","venue":null,"work_id":null,"year":1909},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":47,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.639213Z"},"links":{"cited_paper":"/paper/1909.01496","citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:620666d1082f5b1b8e237aafed2c44c079d7a71894eb49a44fe7f386e1c6193d","observation_id":"50621b13-4dea-4bd8-9302-9485a17d9a5f","resolution":{"observed_at":"2026-08-06T15:53:17.639213Z","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":"2026-08-06T15:53:17.641864Z","title":"@esa (Ref","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":48,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.641864Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:e29c91bdf0242cde5721dd9dded4d776c40e15a3db697725e3c8282d65e5ad61","observation_id":"4e7f7db9-e12e-4296-9d54-e72ff7da2644","resolution":{"observed_at":"2026-08-06T15:53:17.641864Z","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":"2026-08-06T15:53:17.645208Z","title":null,"venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":49,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.645208Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:79802d0a8ade562fbabb4b1103eaffc92664ea53e935eff45f676074fd569963","observation_id":"11e2c744-8834-4d24-88dd-dc9d3978ac8c","resolution":{"observed_at":"2026-08-06T15:53:17.645208Z","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":"2026-08-06T15:53:17.648188Z","title":"teacher” model with some trait T (such as liking owls or being misaligned) generates a dataset consisting solely of number sequences. Remarkably, a “student","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data","version":1},"reference_index":50,"source":"arxiv_source","source_observed_at":"2026-08-06T15:53:17.648188Z"},"links":{"citing_paper":"/paper/2507.14805"},"observation_digest":"sha256:bfd137f73fe2c9e4a24505ce7ee0dcee6c6ab968c7a8c857620b1e89a2f26072","observation_id":"34d1fe3d-07dd-4437-ae51-83a3f1e6b530","resolution":{"observed_at":"2026-08-06T15:53:17.648188Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"paper":{"arxiv_id":"2507.14805","last_updated":"2025-07-20T03:51:13Z","latest_version":1,"primary_category":"cs.LG","snapshot_observed_at":"2026-08-14T23:23:40.747426Z","submitted_at":"2025-07-20T03:51:13Z","title":"Subliminal Learning: Language models transmit behavioral traits via hidden signals in data"},"reference_resolution":{"displayed":50,"state_counts":{"malformed_identifier":0,"metadata_mismatch":0,"parse_uncertain":0,"unresolved":34,"verified_exact":1,"verified_fuzzy":15},"total_outbound_references":50},"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":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"thesis":"As of 19 August 2026, this Paper Citation Record lists 50 of 50 outbound references and 34 inbound Pith citation observations for arXiv:2507.14805."}