{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:Q3CCEY5ZXVZ7K55MMUBIVCCUXB","short_pith_number":"pith:Q3CCEY5Z","canonical_record":{"source":{"id":"2502.11356","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-17T02:11:17Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"bbaf164023d2f8839b2aefe48fbb5a47833b7c730d684cfc8ae55d6760778e92","abstract_canon_sha256":"eac743c167fe4e336e11d4390a06e1c1a02a80b944f8dfcdddc6c37c9c1115ab"},"schema_version":"1.0"},"canonical_sha256":"86c42263b9bd73f577ac65028a8854b865c246aa1b26dbcc8108580c12ec7674","source":{"kind":"arxiv","id":"2502.11356","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.11356","created_at":"2026-07-05T10:15:27Z"},{"alias_kind":"arxiv_version","alias_value":"2502.11356v1","created_at":"2026-07-05T10:15:27Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.11356","created_at":"2026-07-05T10:15:27Z"},{"alias_kind":"pith_short_12","alias_value":"Q3CCEY5ZXVZ7","created_at":"2026-07-05T10:15:27Z"},{"alias_kind":"pith_short_16","alias_value":"Q3CCEY5ZXVZ7K55M","created_at":"2026-07-05T10:15:27Z"},{"alias_kind":"pith_short_8","alias_value":"Q3CCEY5Z","created_at":"2026-07-05T10:15:27Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:Q3CCEY5ZXVZ7K55MMUBIVCCUXB","target":"record","payload":{"canonical_record":{"source":{"id":"2502.11356","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-17T02:11:17Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"bbaf164023d2f8839b2aefe48fbb5a47833b7c730d684cfc8ae55d6760778e92","abstract_canon_sha256":"eac743c167fe4e336e11d4390a06e1c1a02a80b944f8dfcdddc6c37c9c1115ab"},"schema_version":"1.0"},"canonical_sha256":"86c42263b9bd73f577ac65028a8854b865c246aa1b26dbcc8108580c12ec7674","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:15:27.089962Z","signature_b64":"QlHOAvBU6cOIvpKve0FS9xtaPxlqtPhxgUrrzo1aWKgMdpngnaCh0Ge4KPGqqyDzpIH21dtcFjKwUVwwd+FbBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"86c42263b9bd73f577ac65028a8854b865c246aa1b26dbcc8108580c12ec7674","last_reissued_at":"2026-07-05T10:15:27.089473Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:15:27.089473Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2502.11356","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:15:27Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"h9hXgRovkeWqZaq9pCxKZPBr8M7by31292wwdqa0+UITvzwYz/Xx9LnkZjPBB35aTzI1xFPTuKQnuWLlIRH4CQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T18:33:37.820016Z"},"content_sha256":"27d81a692e946f25c4ef22ccf4784e8e64399852ee1bc98c92529ee3379b3dfb","schema_version":"1.0","event_id":"sha256:27d81a692e946f25c4ef22ccf4784e8e64399852ee1bc98c92529ee3379b3dfb"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:Q3CCEY5ZXVZ7K55MMUBIVCCUXB","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"SAIF: A Sparse Autoencoder Framework for Interpreting and Steering Instruction Following of Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Ali Payani, Fan Yang, Haiyan Zhao, Jing Ma, Mengnan Du, Yiran Qiao, Zirui He","submitted_at":"2025-02-17T02:11:17Z","abstract_excerpt":"The ability of large language models (LLMs) to follow instructions is crucial for their practical applications, yet the underlying mechanisms remain poorly understood. This paper presents a novel framework that leverages sparse autoencoders (SAE) to interpret how instruction following works in these models. We demonstrate how the features we identify can effectively steer model outputs to align with given instructions. Through analysis of SAE latent activations, we identify specific latents responsible for instruction following behavior. Our findings reveal that instruction following capabilit"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.11356","kind":"arxiv","version":1},"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/2502.11356/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":"2026-07-05T10:15:27Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LWoNnmby1uJ+GylOrYB3uGHygPTgai731fZlALQWyTYW4uQLdJRk4/9bnIxEITNQ0TUe2TEbU/u5Iur0BstjDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T18:33:37.820528Z"},"content_sha256":"b41d0d981af5dc84928525d0051640a411c3ca982f448ae9f4722c7293afefb4","schema_version":"1.0","event_id":"sha256:b41d0d981af5dc84928525d0051640a411c3ca982f448ae9f4722c7293afefb4"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/Q3CCEY5ZXVZ7K55MMUBIVCCUXB/bundle.json","state_url":"https://pith.science/pith/Q3CCEY5ZXVZ7K55MMUBIVCCUXB/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/Q3CCEY5ZXVZ7K55MMUBIVCCUXB/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":"2026-08-07T18:33:37Z","links":{"resolver":"https://pith.science/pith/Q3CCEY5ZXVZ7K55MMUBIVCCUXB","bundle":"https://pith.science/pith/Q3CCEY5ZXVZ7K55MMUBIVCCUXB/bundle.json","state":"https://pith.science/pith/Q3CCEY5ZXVZ7K55MMUBIVCCUXB/state.json","well_known_bundle":"https://pith.science/.well-known/pith/Q3CCEY5ZXVZ7K55MMUBIVCCUXB/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:Q3CCEY5ZXVZ7K55MMUBIVCCUXB","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":"eac743c167fe4e336e11d4390a06e1c1a02a80b944f8dfcdddc6c37c9c1115ab","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-17T02:11:17Z","title_canon_sha256":"bbaf164023d2f8839b2aefe48fbb5a47833b7c730d684cfc8ae55d6760778e92"},"schema_version":"1.0","source":{"id":"2502.11356","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.11356","created_at":"2026-07-05T10:15:27Z"},{"alias_kind":"arxiv_version","alias_value":"2502.11356v1","created_at":"2026-07-05T10:15:27Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.11356","created_at":"2026-07-05T10:15:27Z"},{"alias_kind":"pith_short_12","alias_value":"Q3CCEY5ZXVZ7","created_at":"2026-07-05T10:15:27Z"},{"alias_kind":"pith_short_16","alias_value":"Q3CCEY5ZXVZ7K55M","created_at":"2026-07-05T10:15:27Z"},{"alias_kind":"pith_short_8","alias_value":"Q3CCEY5Z","created_at":"2026-07-05T10:15:27Z"}],"graph_snapshots":[{"event_id":"sha256:b41d0d981af5dc84928525d0051640a411c3ca982f448ae9f4722c7293afefb4","target":"graph","created_at":"2026-07-05T10:15:27Z","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/2502.11356/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The ability of large language models (LLMs) to follow instructions is crucial for their practical applications, yet the underlying mechanisms remain poorly understood. This paper presents a novel framework that leverages sparse autoencoders (SAE) to interpret how instruction following works in these models. We demonstrate how the features we identify can effectively steer model outputs to align with given instructions. Through analysis of SAE latent activations, we identify specific latents responsible for instruction following behavior. Our findings reveal that instruction following capabilit","authors_text":"Ali Payani, Fan Yang, Haiyan Zhao, Jing Ma, Mengnan Du, Yiran Qiao, Zirui He","cross_cats":["cs.AI","cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-17T02:11:17Z","title":"SAIF: A Sparse Autoencoder Framework for Interpreting and Steering Instruction Following of Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.11356","kind":"arxiv","version":1},"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":"sha256:27d81a692e946f25c4ef22ccf4784e8e64399852ee1bc98c92529ee3379b3dfb","target":"record","created_at":"2026-07-05T10:15:27Z","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":"eac743c167fe4e336e11d4390a06e1c1a02a80b944f8dfcdddc6c37c9c1115ab","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-17T02:11:17Z","title_canon_sha256":"bbaf164023d2f8839b2aefe48fbb5a47833b7c730d684cfc8ae55d6760778e92"},"schema_version":"1.0","source":{"id":"2502.11356","kind":"arxiv","version":1}},"canonical_sha256":"86c42263b9bd73f577ac65028a8854b865c246aa1b26dbcc8108580c12ec7674","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"86c42263b9bd73f577ac65028a8854b865c246aa1b26dbcc8108580c12ec7674","first_computed_at":"2026-07-05T10:15:27.089473Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:15:27.089473Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"QlHOAvBU6cOIvpKve0FS9xtaPxlqtPhxgUrrzo1aWKgMdpngnaCh0Ge4KPGqqyDzpIH21dtcFjKwUVwwd+FbBw==","signature_status":"signed_v1","signed_at":"2026-07-05T10:15:27.089962Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.11356","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:27d81a692e946f25c4ef22ccf4784e8e64399852ee1bc98c92529ee3379b3dfb","sha256:b41d0d981af5dc84928525d0051640a411c3ca982f448ae9f4722c7293afefb4"],"state_sha256":"fcb3fdff1dde6a653881b9d0556136ab678e9e58bbd4b41e79f6da544e702af1"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WDMIpMhjOmgV4BrxVpcRX7xD4krD7DZzUhVz6MW52Eu3bE0ByFONtmzAZBtySfJ7lbeRqp6Q1XM5lKXJ+tiLAw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-07T18:33:37.826141Z","bundle_sha256":"693270dc58f68b54d7d1f10e30c90ef4b1157f07739fdada44561ab9a35b5c90"}}