{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:RADVO4ZGCQO7VFKWNOFW3QRZEV","short_pith_number":"pith:RADVO4ZG","schema_version":"1.0","canonical_sha256":"8807577326141dfa95566b8b6dc2392562d3520316a17d2ff70b9bd2e7b769c0","source":{"kind":"arxiv","id":"2510.09885","version":6},"attestation_state":"computed","paper":{"title":"Diffusion-Inspired Masked Fine-Tuning for Knowledge Injection in Autoregressive LLMs","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Masked fine-tuning lets autoregressive LLMs absorb new facts without paraphrases and without reversal-curse failures.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ely Hahami, Haim Sompolinsky, Jingxuan Fan, Xu Pan, Ziqian Xie","submitted_at":"2025-10-10T21:43:50Z","abstract_excerpt":"Large language models (LLMs) are often used in environments where facts evolve, yet factual knowledge updates via fine-tuning on unstructured text often suffer from 1) reliance on compute-heavy paraphrasing augmentation and 2) the reversal curse. Recent studies show diffusion large language models (dLLMs) require fewer training samples to achieve lower loss in pre-training and are more resistant to the reversal curse, suggesting dLLMs may learn new knowledge more easily than autoregressive LLMs (arLLMs). We test this hypothesis in controlled knowledge fine-tuning experiments and find that whil"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2510.09885","kind":"arxiv","version":6},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2025-10-10T21:43:50Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"ac8e2cde2e0a5981f7ccbd503bdf9f1b6993a0e2a55da1770bad22efc1ee93e2","abstract_canon_sha256":"b54915ca9ea819f2c347239d64dc926c9b8813cb1d7b177bb46ddbe7083d99a2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-11T01:09:19.891532Z","signature_b64":"hU8HAYl1NuvCI+9h9xt0DJfSoG32aiprIuqGs1jjzwS1vZEADF+Uiss/C9/lDiIv6UtZWsVtkA4bzp8QVm04Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8807577326141dfa95566b8b6dc2392562d3520316a17d2ff70b9bd2e7b769c0","last_reissued_at":"2026-06-11T01:09:19.890707Z","signature_status":"signed_v1","first_computed_at":"2026-06-11T01:09:19.890707Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Diffusion-Inspired Masked Fine-Tuning for Knowledge Injection in Autoregressive LLMs","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Masked fine-tuning lets autoregressive LLMs absorb new facts without paraphrases and without reversal-curse failures.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ely Hahami, Haim Sompolinsky, Jingxuan Fan, Xu Pan, Ziqian Xie","submitted_at":"2025-10-10T21:43:50Z","abstract_excerpt":"Large language models (LLMs) are often used in environments where facts evolve, yet factual knowledge updates via fine-tuning on unstructured text often suffer from 1) reliance on compute-heavy paraphrasing augmentation and 2) the reversal curse. Recent studies show diffusion large language models (dLLMs) require fewer training samples to achieve lower loss in pre-training and are more resistant to the reversal curse, suggesting dLLMs may learn new knowledge more easily than autoregressive LLMs (arLLMs). We test this hypothesis in controlled knowledge fine-tuning experiments and find that whil"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The masked fine-tuning for arLLMs substantially improves the efficacy of knowledge injection, i.e. no paraphrase needed and resistant to the reversal curse, closing the gap between arLLMs and dLLMs.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the demasking objective alone induces the observed knowledge-injection advantage in arLLMs independent of diffusion-specific architecture details, and that the controlled experiments isolate this effect without confounding differences in model scale, data distribution, or masking implementation.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Masked fine-tuning enables autoregressive LLMs to inject new factual knowledge without paraphrases and with reversal-curse resistance, matching diffusion LLM advantages on QA tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Masked fine-tuning lets autoregressive LLMs absorb new facts without paraphrases and without reversal-curse failures.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"d9d6daa9402ffd87a5266e86cad8b4d6e476d3026019d589a317f4fe0cdb35bf"},"source":{"id":"2510.09885","kind":"arxiv","version":6},"verdict":{"id":"4305538a-0d9a-4d9c-855d-0ec86d924738","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-18T07:18:49.407331Z","strongest_claim":"The masked fine-tuning for arLLMs substantially improves the efficacy of knowledge injection, i.e. no paraphrase needed and resistant to the reversal curse, closing the gap between arLLMs and dLLMs.","one_line_summary":"Masked fine-tuning enables autoregressive LLMs to inject new factual knowledge without paraphrases and with reversal-curse resistance, matching diffusion LLM advantages on QA tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the demasking objective alone induces the observed knowledge-injection advantage in arLLMs independent of diffusion-specific architecture details, and that the controlled experiments isolate this effect without confounding differences in model scale, data distribution, or masking implementation.","pith_extraction_headline":"Masked fine-tuning lets autoregressive LLMs absorb new facts without paraphrases and without reversal-curse failures."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2510.09885/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2510.09885","created_at":"2026-06-11T01:09:19.890842+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.09885v6","created_at":"2026-06-11T01:09:19.890842+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.09885","created_at":"2026-06-11T01:09:19.890842+00:00"},{"alias_kind":"pith_short_12","alias_value":"RADVO4ZGCQO7","created_at":"2026-06-11T01:09:19.890842+00:00"},{"alias_kind":"pith_short_16","alias_value":"RADVO4ZGCQO7VFKW","created_at":"2026-06-11T01:09:19.890842+00:00"},{"alias_kind":"pith_short_8","alias_value":"RADVO4ZG","created_at":"2026-06-11T01:09:19.890842+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2606.03924","citing_title":"Knowledge Editing in Masked Diffusion Language Models","ref_index":57,"is_internal_anchor":true},{"citing_arxiv_id":"2604.04943","citing_title":"The Illusion of Latent Generalization: Bi-directionality and the Reversal Curse","ref_index":7,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RADVO4ZGCQO7VFKWNOFW3QRZEV","json":"https://pith.science/pith/RADVO4ZGCQO7VFKWNOFW3QRZEV.json","graph_json":"https://pith.science/api/pith-number/RADVO4ZGCQO7VFKWNOFW3QRZEV/graph.json","events_json":"https://pith.science/api/pith-number/RADVO4ZGCQO7VFKWNOFW3QRZEV/events.json","paper":"https://pith.science/paper/RADVO4ZG"},"agent_actions":{"view_html":"https://pith.science/pith/RADVO4ZGCQO7VFKWNOFW3QRZEV","download_json":"https://pith.science/pith/RADVO4ZGCQO7VFKWNOFW3QRZEV.json","view_paper":"https://pith.science/paper/RADVO4ZG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.09885&json=true","fetch_graph":"https://pith.science/api/pith-number/RADVO4ZGCQO7VFKWNOFW3QRZEV/graph.json","fetch_events":"https://pith.science/api/pith-number/RADVO4ZGCQO7VFKWNOFW3QRZEV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RADVO4ZGCQO7VFKWNOFW3QRZEV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RADVO4ZGCQO7VFKWNOFW3QRZEV/action/storage_attestation","attest_author":"https://pith.science/pith/RADVO4ZGCQO7VFKWNOFW3QRZEV/action/author_attestation","sign_citation":"https://pith.science/pith/RADVO4ZGCQO7VFKWNOFW3QRZEV/action/citation_signature","submit_replication":"https://pith.science/pith/RADVO4ZGCQO7VFKWNOFW3QRZEV/action/replication_record"}},"created_at":"2026-06-11T01:09:19.890842+00:00","updated_at":"2026-06-11T01:09:19.890842+00:00"}