{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:5JIMKZOKA7UANWYSW6LHPQTMSD","short_pith_number":"pith:5JIMKZOK","canonical_record":{"source":{"id":"2411.07618","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-11-12T07:54:13Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"0fd9c99c6f88b680f444fdf1b73acc497f3adc0c2206780e011d1c1c82058416","abstract_canon_sha256":"f14445454e47ee350331a4031e8d8739eeb99e5610287260f1b3a5d32a7f8dc7"},"schema_version":"1.0"},"canonical_sha256":"ea50c565ca07e806db12b79677c26c90e09924492d80910ac3bba3019e83ea5d","source":{"kind":"arxiv","id":"2411.07618","version":4},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2411.07618","created_at":"2026-07-05T11:34:44Z"},{"alias_kind":"arxiv_version","alias_value":"2411.07618v4","created_at":"2026-07-05T11:34:44Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.07618","created_at":"2026-07-05T11:34:44Z"},{"alias_kind":"pith_short_12","alias_value":"5JIMKZOKA7UA","created_at":"2026-07-05T11:34:44Z"},{"alias_kind":"pith_short_16","alias_value":"5JIMKZOKA7UANWYS","created_at":"2026-07-05T11:34:44Z"},{"alias_kind":"pith_short_8","alias_value":"5JIMKZOK","created_at":"2026-07-05T11:34:44Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:5JIMKZOKA7UANWYSW6LHPQTMSD","target":"record","payload":{"canonical_record":{"source":{"id":"2411.07618","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-11-12T07:54:13Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"0fd9c99c6f88b680f444fdf1b73acc497f3adc0c2206780e011d1c1c82058416","abstract_canon_sha256":"f14445454e47ee350331a4031e8d8739eeb99e5610287260f1b3a5d32a7f8dc7"},"schema_version":"1.0"},"canonical_sha256":"ea50c565ca07e806db12b79677c26c90e09924492d80910ac3bba3019e83ea5d","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:34:44.941823Z","signature_b64":"pt95M+Zri/Qj/ibnSt/LtT1nBIALU59RqCBG7sLkrxnvdPU6W/OBxt2RJ0TLbjII+FBE7Mdoxz4GR5NZmjh5Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ea50c565ca07e806db12b79677c26c90e09924492d80910ac3bba3019e83ea5d","last_reissued_at":"2026-07-05T11:34:44.941275Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:34:44.941275Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2411.07618","source_version":4,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:34:44Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6uhkrtJBc9Rqkgvr+ANgZafDumECkO10LMJ5AtmJFzTS/Fsikfwp8639yxkLG9JaKo68XpxLAVgs3ZXwqskgAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T00:16:16.277672Z"},"content_sha256":"3c3bcba990a8a8d4bbcd8711f89099011dc9cc4e110c3248a736803255f4066c","schema_version":"1.0","event_id":"sha256:3c3bcba990a8a8d4bbcd8711f89099011dc9cc4e110c3248a736803255f4066c"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:5JIMKZOKA7UANWYSW6LHPQTMSD","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Constrain Alignment with Sparse Autoencoders","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Chak Tou Leong, Hanqi Yan, Jun Wang, Linyi Yang, Minjun Zhu, Qiang Zhang, Qingyu Yin, Wenjie Li, Yue Zhang, Yulan He","submitted_at":"2024-11-12T07:54:13Z","abstract_excerpt":"The alignment of large language models (LLMs) with human preferences remains a key challenge. While post-training techniques like Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO) have achieved notable success, they often introduce computational inefficiencies and training instability. In this paper, we propose Feature-level constrained Preference Optimization (FPO), a novel method designed to simplify the alignment process while ensuring stability. FPO leverages pre-trained Sparse Autoencoders (SAEs) and introduces feature-level constraints, allowing f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.07618","kind":"arxiv","version":4},"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/2411.07618/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":"2026-07-05T11:34:44Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LgM8hpxIibosG/FOZbayN8M/d6XZEzz/ybLxqzLBuKT6WJiJdc6WjhRIzOFHaINYfGFhf9OIG0iLvVPXaB5aCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T00:16:16.278386Z"},"content_sha256":"b327f4b62296aa691d17323e54bcae146d57d92728355f57d201ae1458e4053d","schema_version":"1.0","event_id":"sha256:b327f4b62296aa691d17323e54bcae146d57d92728355f57d201ae1458e4053d"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/5JIMKZOKA7UANWYSW6LHPQTMSD/bundle.json","state_url":"https://pith.science/pith/5JIMKZOKA7UANWYSW6LHPQTMSD/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/5JIMKZOKA7UANWYSW6LHPQTMSD/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":"2026-08-05T00:16:16Z","links":{"resolver":"https://pith.science/pith/5JIMKZOKA7UANWYSW6LHPQTMSD","bundle":"https://pith.science/pith/5JIMKZOKA7UANWYSW6LHPQTMSD/bundle.json","state":"https://pith.science/pith/5JIMKZOKA7UANWYSW6LHPQTMSD/state.json","well_known_bundle":"https://pith.science/.well-known/pith/5JIMKZOKA7UANWYSW6LHPQTMSD/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:5JIMKZOKA7UANWYSW6LHPQTMSD","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":"f14445454e47ee350331a4031e8d8739eeb99e5610287260f1b3a5d32a7f8dc7","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-11-12T07:54:13Z","title_canon_sha256":"0fd9c99c6f88b680f444fdf1b73acc497f3adc0c2206780e011d1c1c82058416"},"schema_version":"1.0","source":{"id":"2411.07618","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2411.07618","created_at":"2026-07-05T11:34:44Z"},{"alias_kind":"arxiv_version","alias_value":"2411.07618v4","created_at":"2026-07-05T11:34:44Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.07618","created_at":"2026-07-05T11:34:44Z"},{"alias_kind":"pith_short_12","alias_value":"5JIMKZOKA7UA","created_at":"2026-07-05T11:34:44Z"},{"alias_kind":"pith_short_16","alias_value":"5JIMKZOKA7UANWYS","created_at":"2026-07-05T11:34:44Z"},{"alias_kind":"pith_short_8","alias_value":"5JIMKZOK","created_at":"2026-07-05T11:34:44Z"}],"graph_snapshots":[{"event_id":"sha256:b327f4b62296aa691d17323e54bcae146d57d92728355f57d201ae1458e4053d","target":"graph","created_at":"2026-07-05T11:34:44Z","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/2411.07618/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The alignment of large language models (LLMs) with human preferences remains a key challenge. While post-training techniques like Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO) have achieved notable success, they often introduce computational inefficiencies and training instability. In this paper, we propose Feature-level constrained Preference Optimization (FPO), a novel method designed to simplify the alignment process while ensuring stability. FPO leverages pre-trained Sparse Autoencoders (SAEs) and introduces feature-level constraints, allowing f","authors_text":"Chak Tou Leong, Hanqi Yan, Jun Wang, Linyi Yang, Minjun Zhu, Qiang Zhang, Qingyu Yin, Wenjie Li, Yue Zhang, Yulan He","cross_cats":["cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-11-12T07:54:13Z","title":"Constrain Alignment with Sparse Autoencoders"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.07618","kind":"arxiv","version":4},"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":"sha256:3c3bcba990a8a8d4bbcd8711f89099011dc9cc4e110c3248a736803255f4066c","target":"record","created_at":"2026-07-05T11:34:44Z","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":"f14445454e47ee350331a4031e8d8739eeb99e5610287260f1b3a5d32a7f8dc7","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-11-12T07:54:13Z","title_canon_sha256":"0fd9c99c6f88b680f444fdf1b73acc497f3adc0c2206780e011d1c1c82058416"},"schema_version":"1.0","source":{"id":"2411.07618","kind":"arxiv","version":4}},"canonical_sha256":"ea50c565ca07e806db12b79677c26c90e09924492d80910ac3bba3019e83ea5d","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"ea50c565ca07e806db12b79677c26c90e09924492d80910ac3bba3019e83ea5d","first_computed_at":"2026-07-05T11:34:44.941275Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:34:44.941275Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"pt95M+Zri/Qj/ibnSt/LtT1nBIALU59RqCBG7sLkrxnvdPU6W/OBxt2RJ0TLbjII+FBE7Mdoxz4GR5NZmjh5Aw==","signature_status":"signed_v1","signed_at":"2026-07-05T11:34:44.941823Z","signed_message":"canonical_sha256_bytes"},"source_id":"2411.07618","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:3c3bcba990a8a8d4bbcd8711f89099011dc9cc4e110c3248a736803255f4066c","sha256:b327f4b62296aa691d17323e54bcae146d57d92728355f57d201ae1458e4053d"],"state_sha256":"8ab1e36ad26e10c477250d58b27a2c381a944f53935d24df3047306df3439419"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ID1JLAnCSqzZZRRkhHfGzAg4uTnQ5mWGQi5WPTqoH/8MromstOeLir9KmSO9qHWqCkAR6FDl3SSvt30fcY8KCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T00:16:16.288927Z","bundle_sha256":"5c75e1a94117a845ea97e6b74a384ffc86e072c8d072f350fea32d136bb4b7b8"}}