{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:IVDSZZHZZWGYDEWCGZNB7PCK72","short_pith_number":"pith:IVDSZZHZ","canonical_record":{"source":{"id":"2305.17333","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-27T02:28:10Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"0cd953b312c8c193f30a98a16e06f292a88df6c34e777e39a4c35b02fe4ed61a","abstract_canon_sha256":"c0595438410fbafc0d055e338ece5d25bf2a633a11afb1b111af03a1175bc678"},"schema_version":"1.0"},"canonical_sha256":"45472ce4f9cd8d8192c2365a1fbc4afe8d400a1e21e8e37080af6e7ef271adac","source":{"kind":"arxiv","id":"2305.17333","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2305.17333","created_at":"2026-07-05T07:32:18Z"},{"alias_kind":"arxiv_version","alias_value":"2305.17333v3","created_at":"2026-07-05T07:32:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.17333","created_at":"2026-07-05T07:32:18Z"},{"alias_kind":"pith_short_12","alias_value":"IVDSZZHZZWGY","created_at":"2026-07-05T07:32:18Z"},{"alias_kind":"pith_short_16","alias_value":"IVDSZZHZZWGYDEWC","created_at":"2026-07-05T07:32:18Z"},{"alias_kind":"pith_short_8","alias_value":"IVDSZZHZ","created_at":"2026-07-05T07:32:18Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:IVDSZZHZZWGYDEWCGZNB7PCK72","target":"record","payload":{"canonical_record":{"source":{"id":"2305.17333","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-27T02:28:10Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"0cd953b312c8c193f30a98a16e06f292a88df6c34e777e39a4c35b02fe4ed61a","abstract_canon_sha256":"c0595438410fbafc0d055e338ece5d25bf2a633a11afb1b111af03a1175bc678"},"schema_version":"1.0"},"canonical_sha256":"45472ce4f9cd8d8192c2365a1fbc4afe8d400a1e21e8e37080af6e7ef271adac","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:32:18.772207Z","signature_b64":"I6pebu+AKXN/HmbGt9Vg1/zGTaTN85zRIKw10xclT+flAxIq2nk3RRGLvee8DCg6cxgmHk7LszswKFi4kZUOAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"45472ce4f9cd8d8192c2365a1fbc4afe8d400a1e21e8e37080af6e7ef271adac","last_reissued_at":"2026-07-05T07:32:18.771750Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:32:18.771750Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2305.17333","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:32:18Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"GirWq2Fo/5G87+DPzxGxiTOEHXVHMtNfQhoQzq6Z2s6QeyCtnu2FKOPX4hGCmwfyu0GEN8pTyTH0he83G+2HDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-16T04:18:58.390011Z"},"content_sha256":"369db4f1b7dea49b2895d7a94a57a57c7e22c20578f509771bf3d3889e258ab0","schema_version":"1.0","event_id":"sha256:369db4f1b7dea49b2895d7a94a57a57c7e22c20578f509771bf3d3889e258ab0"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:IVDSZZHZZWGYDEWCGZNB7PCK72","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Fine-Tuning Language Models with Just Forward Passes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Alex Damian, Danqi Chen, Eshaan Nichani, Jason D. Lee, Sadhika Malladi, Sanjeev Arora, Tianyu Gao","submitted_at":"2023-05-27T02:28:10Z","abstract_excerpt":"Fine-tuning language models (LMs) has yielded success on diverse downstream tasks, but as LMs grow in size, backpropagation requires a prohibitively large amount of memory. Zeroth-order (ZO) methods can in principle estimate gradients using only two forward passes but are theorized to be catastrophically slow for optimizing large models. In this work, we propose a memory-efficient zerothorder optimizer (MeZO), adapting the classical ZO-SGD method to operate in-place, thereby fine-tuning LMs with the same memory footprint as inference. For example, with a single A100 80GB GPU, MeZO can train a "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.17333","kind":"arxiv","version":3},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2305.17333/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:32:18Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XAyrlT6v8OB7QyT1ZJz+8Gyl4DujJRAMiekLcJ2CcCqeOU6EiLzzUrgIRsFV1bju8b9QLzBxAbzC6DBGUZIHBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-16T04:18:58.390503Z"},"content_sha256":"a23d957cb6612162b0c93a8d78b71aea917b4188fc80481cc180240cbe632180","schema_version":"1.0","event_id":"sha256:a23d957cb6612162b0c93a8d78b71aea917b4188fc80481cc180240cbe632180"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/IVDSZZHZZWGYDEWCGZNB7PCK72/bundle.json","state_url":"https://pith.science/pith/IVDSZZHZZWGYDEWCGZNB7PCK72/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/IVDSZZHZZWGYDEWCGZNB7PCK72/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-16T04:18:58Z","links":{"resolver":"https://pith.science/pith/IVDSZZHZZWGYDEWCGZNB7PCK72","bundle":"https://pith.science/pith/IVDSZZHZZWGYDEWCGZNB7PCK72/bundle.json","state":"https://pith.science/pith/IVDSZZHZZWGYDEWCGZNB7PCK72/state.json","well_known_bundle":"https://pith.science/.well-known/pith/IVDSZZHZZWGYDEWCGZNB7PCK72/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:IVDSZZHZZWGYDEWCGZNB7PCK72","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"c0595438410fbafc0d055e338ece5d25bf2a633a11afb1b111af03a1175bc678","cross_cats_sorted":["cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-27T02:28:10Z","title_canon_sha256":"0cd953b312c8c193f30a98a16e06f292a88df6c34e777e39a4c35b02fe4ed61a"},"schema_version":"1.0","source":{"id":"2305.17333","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2305.17333","created_at":"2026-07-05T07:32:18Z"},{"alias_kind":"arxiv_version","alias_value":"2305.17333v3","created_at":"2026-07-05T07:32:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.17333","created_at":"2026-07-05T07:32:18Z"},{"alias_kind":"pith_short_12","alias_value":"IVDSZZHZZWGY","created_at":"2026-07-05T07:32:18Z"},{"alias_kind":"pith_short_16","alias_value":"IVDSZZHZZWGYDEWC","created_at":"2026-07-05T07:32:18Z"},{"alias_kind":"pith_short_8","alias_value":"IVDSZZHZ","created_at":"2026-07-05T07:32:18Z"}],"graph_snapshots":[{"event_id":"sha256:a23d957cb6612162b0c93a8d78b71aea917b4188fc80481cc180240cbe632180","target":"graph","created_at":"2026-07-05T07:32:18Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2305.17333/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Fine-tuning language models (LMs) has yielded success on diverse downstream tasks, but as LMs grow in size, backpropagation requires a prohibitively large amount of memory. Zeroth-order (ZO) methods can in principle estimate gradients using only two forward passes but are theorized to be catastrophically slow for optimizing large models. In this work, we propose a memory-efficient zerothorder optimizer (MeZO), adapting the classical ZO-SGD method to operate in-place, thereby fine-tuning LMs with the same memory footprint as inference. For example, with a single A100 80GB GPU, MeZO can train a ","authors_text":"Alex Damian, Danqi Chen, Eshaan Nichani, Jason D. Lee, Sadhika Malladi, Sanjeev Arora, Tianyu Gao","cross_cats":["cs.CL"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-27T02:28:10Z","title":"Fine-Tuning Language Models with Just Forward Passes"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.17333","kind":"arxiv","version":3},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:369db4f1b7dea49b2895d7a94a57a57c7e22c20578f509771bf3d3889e258ab0","target":"record","created_at":"2026-07-05T07:32:18Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"c0595438410fbafc0d055e338ece5d25bf2a633a11afb1b111af03a1175bc678","cross_cats_sorted":["cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-27T02:28:10Z","title_canon_sha256":"0cd953b312c8c193f30a98a16e06f292a88df6c34e777e39a4c35b02fe4ed61a"},"schema_version":"1.0","source":{"id":"2305.17333","kind":"arxiv","version":3}},"canonical_sha256":"45472ce4f9cd8d8192c2365a1fbc4afe8d400a1e21e8e37080af6e7ef271adac","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"45472ce4f9cd8d8192c2365a1fbc4afe8d400a1e21e8e37080af6e7ef271adac","first_computed_at":"2026-07-05T07:32:18.771750Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T07:32:18.771750Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"I6pebu+AKXN/HmbGt9Vg1/zGTaTN85zRIKw10xclT+flAxIq2nk3RRGLvee8DCg6cxgmHk7LszswKFi4kZUOAQ==","signature_status":"signed_v1","signed_at":"2026-07-05T07:32:18.772207Z","signed_message":"canonical_sha256_bytes"},"source_id":"2305.17333","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:369db4f1b7dea49b2895d7a94a57a57c7e22c20578f509771bf3d3889e258ab0","sha256:a23d957cb6612162b0c93a8d78b71aea917b4188fc80481cc180240cbe632180"],"state_sha256":"89461b8af070cad338470e1d7c80c774279e9ef78542b7dc187d6e29d8e33d63"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"d2QatcXegjTXfbMw35T+mXZonUlMc0/FBThuJnIIt/3SefCpdw+Z3wLzSwCsiMgo2XKCUv2jTlSEtTUfK6odCQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-16T04:18:58.395800Z","bundle_sha256":"8d0da85b42377e7513c99983380e723af5864bba89ce71e99aa90b3834e1f550"}}