{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:FZDSHQFGGPNWHAY6VGRW2DSISR","short_pith_number":"pith:FZDSHQFG","canonical_record":{"source":{"id":"2412.00967","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-12-01T21:11:28Z","cross_cats_sorted":[],"title_canon_sha256":"e34d789b78103c8cb602fcf68b112b3240875ae9cfa77b6da8aea66fbc47cdde","abstract_canon_sha256":"697a933aa1137ceab0f378fb1a98c3d5ace67d844f26caf2518d629f672becc1"},"schema_version":"1.0"},"canonical_sha256":"2e4723c0a633db63831ea9a36d0e489449da2a3f40250fc8a06968120da7a34c","source":{"kind":"arxiv","id":"2412.00967","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2412.00967","created_at":"2026-07-05T09:42:51Z"},{"alias_kind":"arxiv_version","alias_value":"2412.00967v1","created_at":"2026-07-05T09:42:51Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.00967","created_at":"2026-07-05T09:42:51Z"},{"alias_kind":"pith_short_12","alias_value":"FZDSHQFGGPNW","created_at":"2026-07-05T09:42:51Z"},{"alias_kind":"pith_short_16","alias_value":"FZDSHQFGGPNWHAY6","created_at":"2026-07-05T09:42:51Z"},{"alias_kind":"pith_short_8","alias_value":"FZDSHQFG","created_at":"2026-07-05T09:42:51Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:FZDSHQFGGPNWHAY6VGRW2DSISR","target":"record","payload":{"canonical_record":{"source":{"id":"2412.00967","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-12-01T21:11:28Z","cross_cats_sorted":[],"title_canon_sha256":"e34d789b78103c8cb602fcf68b112b3240875ae9cfa77b6da8aea66fbc47cdde","abstract_canon_sha256":"697a933aa1137ceab0f378fb1a98c3d5ace67d844f26caf2518d629f672becc1"},"schema_version":"1.0"},"canonical_sha256":"2e4723c0a633db63831ea9a36d0e489449da2a3f40250fc8a06968120da7a34c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:42:51.308611Z","signature_b64":"REFX/fnRMj7/tUtZtmeU8j+Ag5kQGmv83i6uYUK780ysak+/sD9/0e3yRbt5YJZUBmyIKJ9Z+BL4yk7VB5WrBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2e4723c0a633db63831ea9a36d0e489449da2a3f40250fc8a06968120da7a34c","last_reissued_at":"2026-07-05T09:42:51.308069Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:42:51.308069Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2412.00967","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":"2026-07-05T09:42:51Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"CQyQK8D8M8iGCAsbxBncn6394pR+jM7ZxbWiEAGAmPzo04GpZ38f59BsqVoJL8PGOUnWYxuLiS9FqDfqbwnYCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T02:38:25.096515Z"},"content_sha256":"d3d315fccd6912a00441ce8a8610e256b0e9a80ced86c585e31ddb4f45cac4d8","schema_version":"1.0","event_id":"sha256:d3d315fccd6912a00441ce8a8610e256b0e9a80ced86c585e31ddb4f45cac4d8"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:FZDSHQFGGPNWHAY6VGRW2DSISR","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Linear Probe Penalties Reduce LLM Sycophancy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Henry Papadatos, Rachel Freedman","submitted_at":"2024-12-01T21:11:28Z","abstract_excerpt":"Large language models (LLMs) are often sycophantic, prioritizing agreement with their users over accurate or objective statements. This problematic behavior becomes more pronounced during reinforcement learning from human feedback (RLHF), an LLM fine-tuning stage intended to align model outputs with human values. Instead of increasing accuracy and reliability, the reward model learned from RLHF often rewards sycophancy. We develop a linear probing method to identify and penalize markers of sycophancy within the reward model, producing rewards that discourage sycophantic behavior. Our experimen"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.00967","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/2412.00967/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":"2026-07-05T09:42:51Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"T0t/scyLxjsWxO1LA01kj0ALIx8V5ZhxNKgWI8NdKBU7bIfbOXNu3pAQUlhCdnIaiQ6ORIAIaOWfe4xDHlPcDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T02:38:25.097154Z"},"content_sha256":"5fc9f1851bdb35783802fe3e95c55620101b2be68eb3b4a989db2525774cf249","schema_version":"1.0","event_id":"sha256:5fc9f1851bdb35783802fe3e95c55620101b2be68eb3b4a989db2525774cf249"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/FZDSHQFGGPNWHAY6VGRW2DSISR/bundle.json","state_url":"https://pith.science/pith/FZDSHQFGGPNWHAY6VGRW2DSISR/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/FZDSHQFGGPNWHAY6VGRW2DSISR/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":"2026-08-04T02:38:25Z","links":{"resolver":"https://pith.science/pith/FZDSHQFGGPNWHAY6VGRW2DSISR","bundle":"https://pith.science/pith/FZDSHQFGGPNWHAY6VGRW2DSISR/bundle.json","state":"https://pith.science/pith/FZDSHQFGGPNWHAY6VGRW2DSISR/state.json","well_known_bundle":"https://pith.science/.well-known/pith/FZDSHQFGGPNWHAY6VGRW2DSISR/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:FZDSHQFGGPNWHAY6VGRW2DSISR","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":"697a933aa1137ceab0f378fb1a98c3d5ace67d844f26caf2518d629f672becc1","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-12-01T21:11:28Z","title_canon_sha256":"e34d789b78103c8cb602fcf68b112b3240875ae9cfa77b6da8aea66fbc47cdde"},"schema_version":"1.0","source":{"id":"2412.00967","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2412.00967","created_at":"2026-07-05T09:42:51Z"},{"alias_kind":"arxiv_version","alias_value":"2412.00967v1","created_at":"2026-07-05T09:42:51Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.00967","created_at":"2026-07-05T09:42:51Z"},{"alias_kind":"pith_short_12","alias_value":"FZDSHQFGGPNW","created_at":"2026-07-05T09:42:51Z"},{"alias_kind":"pith_short_16","alias_value":"FZDSHQFGGPNWHAY6","created_at":"2026-07-05T09:42:51Z"},{"alias_kind":"pith_short_8","alias_value":"FZDSHQFG","created_at":"2026-07-05T09:42:51Z"}],"graph_snapshots":[{"event_id":"sha256:5fc9f1851bdb35783802fe3e95c55620101b2be68eb3b4a989db2525774cf249","target":"graph","created_at":"2026-07-05T09:42:51Z","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/2412.00967/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large language models (LLMs) are often sycophantic, prioritizing agreement with their users over accurate or objective statements. This problematic behavior becomes more pronounced during reinforcement learning from human feedback (RLHF), an LLM fine-tuning stage intended to align model outputs with human values. Instead of increasing accuracy and reliability, the reward model learned from RLHF often rewards sycophancy. We develop a linear probing method to identify and penalize markers of sycophancy within the reward model, producing rewards that discourage sycophantic behavior. Our experimen","authors_text":"Henry Papadatos, Rachel Freedman","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-12-01T21:11:28Z","title":"Linear Probe Penalties Reduce LLM Sycophancy"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.00967","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:d3d315fccd6912a00441ce8a8610e256b0e9a80ced86c585e31ddb4f45cac4d8","target":"record","created_at":"2026-07-05T09:42:51Z","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":"697a933aa1137ceab0f378fb1a98c3d5ace67d844f26caf2518d629f672becc1","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-12-01T21:11:28Z","title_canon_sha256":"e34d789b78103c8cb602fcf68b112b3240875ae9cfa77b6da8aea66fbc47cdde"},"schema_version":"1.0","source":{"id":"2412.00967","kind":"arxiv","version":1}},"canonical_sha256":"2e4723c0a633db63831ea9a36d0e489449da2a3f40250fc8a06968120da7a34c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2e4723c0a633db63831ea9a36d0e489449da2a3f40250fc8a06968120da7a34c","first_computed_at":"2026-07-05T09:42:51.308069Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:42:51.308069Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"REFX/fnRMj7/tUtZtmeU8j+Ag5kQGmv83i6uYUK780ysak+/sD9/0e3yRbt5YJZUBmyIKJ9Z+BL4yk7VB5WrBw==","signature_status":"signed_v1","signed_at":"2026-07-05T09:42:51.308611Z","signed_message":"canonical_sha256_bytes"},"source_id":"2412.00967","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d3d315fccd6912a00441ce8a8610e256b0e9a80ced86c585e31ddb4f45cac4d8","sha256:5fc9f1851bdb35783802fe3e95c55620101b2be68eb3b4a989db2525774cf249"],"state_sha256":"57ea754ba2eabbfb2e48197094f7c73a13c0b0f8927c07f0ddc7d0dc2518cee4"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"G3ih2RZMzbigsapA+jbDzA00c6/lprDGPXUMkUBICUXYXvvnGZsbEuvI01/tjNh27loaY12z6OBeGARXSbkCBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T02:38:25.101164Z","bundle_sha256":"2b77a5e918b32ddb81c4a715d963e3c408501088cfd2dc65142b77b361b5d8c9"}}