{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:T2T7CSVU64HS3F7MZ3I7WFZLM4","short_pith_number":"pith:T2T7CSVU","canonical_record":{"source":{"id":"2409.12822","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-09-19T14:50:34Z","cross_cats_sorted":[],"title_canon_sha256":"4aec86aaca2d9608eb4014a0a9251f32afbfc7bbb3f7f68b06976e7cf9813d0f","abstract_canon_sha256":"40646969e26c3483a945a8a39d02875bab5684919d47deed1671d1fe611b0712"},"schema_version":"1.0"},"canonical_sha256":"9ea7f14ab4f70f2d97ecced1fb172b67301e7564e73f410dde1f66147b5a8b03","source":{"kind":"arxiv","id":"2409.12822","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2409.12822","created_at":"2026-07-05T09:45:54Z"},{"alias_kind":"arxiv_version","alias_value":"2409.12822v3","created_at":"2026-07-05T09:45:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.12822","created_at":"2026-07-05T09:45:54Z"},{"alias_kind":"pith_short_12","alias_value":"T2T7CSVU64HS","created_at":"2026-07-05T09:45:54Z"},{"alias_kind":"pith_short_16","alias_value":"T2T7CSVU64HS3F7M","created_at":"2026-07-05T09:45:54Z"},{"alias_kind":"pith_short_8","alias_value":"T2T7CSVU","created_at":"2026-07-05T09:45:54Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:T2T7CSVU64HS3F7MZ3I7WFZLM4","target":"record","payload":{"canonical_record":{"source":{"id":"2409.12822","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-09-19T14:50:34Z","cross_cats_sorted":[],"title_canon_sha256":"4aec86aaca2d9608eb4014a0a9251f32afbfc7bbb3f7f68b06976e7cf9813d0f","abstract_canon_sha256":"40646969e26c3483a945a8a39d02875bab5684919d47deed1671d1fe611b0712"},"schema_version":"1.0"},"canonical_sha256":"9ea7f14ab4f70f2d97ecced1fb172b67301e7564e73f410dde1f66147b5a8b03","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:45:54.959943Z","signature_b64":"3VsPeseANPyZRf8CJ9/CTvApY2Gx+SwVWzqxYnvNyIeupMroEXXB+UXAaS+rJ1Cv6FArPS5GeNN8+npqdRdvAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9ea7f14ab4f70f2d97ecced1fb172b67301e7564e73f410dde1f66147b5a8b03","last_reissued_at":"2026-07-05T09:45:54.959475Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:45:54.959475Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2409.12822","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":"2026-07-05T09:45:54Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"c7Xqb0toeoH9rlbWQlJoXOKEE3btZnzcTLhIL8WG6N25Jn6ie150XLVpP1pYG72LTFQvlpUk1dinSQVBJo2OBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-13T18:36:14.497518Z"},"content_sha256":"727834885552b1517305fecc940f1a3e4b427966b80587b5d1560b62d37dc236","schema_version":"1.0","event_id":"sha256:727834885552b1517305fecc940f1a3e4b427966b80587b5d1560b62d37dc236"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:T2T7CSVU64HS3F7MZ3I7WFZLM4","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Language Models Learn to Mislead Humans via RLHF","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Akbir Khan, Ethan Perez, He He, Jacob Steinhardt, Jiaxin Wen, Minlie Huang, Ruiqi Zhong, Samuel R. Bowman, Shi Feng","submitted_at":"2024-09-19T14:50:34Z","abstract_excerpt":"Language models (LMs) can produce errors that are hard to detect for humans, especially when the task is complex. RLHF, the most popular post-training method, may exacerbate this problem: to achieve higher rewards, LMs might get better at convincing humans that they are right even when they are wrong. We study this phenomenon under a standard RLHF pipeline, calling it \"U-SOPHISTRY\" since it is Unintended by model developers. Specifically, we ask time-constrained (e.g., 3-10 minutes) human subjects to evaluate the correctness of model outputs and calculate humans' accuracy against gold labels. "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.12822","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/2409.12822/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":"2026-07-05T09:45:54Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"QS3KiB/IjsnaBZYsvZVHNZZMcYmDf4PY+FN7drctrLeRZjHz/cosje0AIJrGHY5vFQj49D7JteGQRtAFt4DvBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-13T18:36:14.497888Z"},"content_sha256":"6ff0de7a808a0ddd2d4eb497abfe8176133d0223ced02384bebc4fcb7b53b527","schema_version":"1.0","event_id":"sha256:6ff0de7a808a0ddd2d4eb497abfe8176133d0223ced02384bebc4fcb7b53b527"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/T2T7CSVU64HS3F7MZ3I7WFZLM4/bundle.json","state_url":"https://pith.science/pith/T2T7CSVU64HS3F7MZ3I7WFZLM4/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/T2T7CSVU64HS3F7MZ3I7WFZLM4/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":"2026-07-13T18:36:14Z","links":{"resolver":"https://pith.science/pith/T2T7CSVU64HS3F7MZ3I7WFZLM4","bundle":"https://pith.science/pith/T2T7CSVU64HS3F7MZ3I7WFZLM4/bundle.json","state":"https://pith.science/pith/T2T7CSVU64HS3F7MZ3I7WFZLM4/state.json","well_known_bundle":"https://pith.science/.well-known/pith/T2T7CSVU64HS3F7MZ3I7WFZLM4/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:T2T7CSVU64HS3F7MZ3I7WFZLM4","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":"40646969e26c3483a945a8a39d02875bab5684919d47deed1671d1fe611b0712","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-09-19T14:50:34Z","title_canon_sha256":"4aec86aaca2d9608eb4014a0a9251f32afbfc7bbb3f7f68b06976e7cf9813d0f"},"schema_version":"1.0","source":{"id":"2409.12822","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2409.12822","created_at":"2026-07-05T09:45:54Z"},{"alias_kind":"arxiv_version","alias_value":"2409.12822v3","created_at":"2026-07-05T09:45:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.12822","created_at":"2026-07-05T09:45:54Z"},{"alias_kind":"pith_short_12","alias_value":"T2T7CSVU64HS","created_at":"2026-07-05T09:45:54Z"},{"alias_kind":"pith_short_16","alias_value":"T2T7CSVU64HS3F7M","created_at":"2026-07-05T09:45:54Z"},{"alias_kind":"pith_short_8","alias_value":"T2T7CSVU","created_at":"2026-07-05T09:45:54Z"}],"graph_snapshots":[{"event_id":"sha256:6ff0de7a808a0ddd2d4eb497abfe8176133d0223ced02384bebc4fcb7b53b527","target":"graph","created_at":"2026-07-05T09:45:54Z","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/2409.12822/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Language models (LMs) can produce errors that are hard to detect for humans, especially when the task is complex. RLHF, the most popular post-training method, may exacerbate this problem: to achieve higher rewards, LMs might get better at convincing humans that they are right even when they are wrong. We study this phenomenon under a standard RLHF pipeline, calling it \"U-SOPHISTRY\" since it is Unintended by model developers. Specifically, we ask time-constrained (e.g., 3-10 minutes) human subjects to evaluate the correctness of model outputs and calculate humans' accuracy against gold labels. ","authors_text":"Akbir Khan, Ethan Perez, He He, Jacob Steinhardt, Jiaxin Wen, Minlie Huang, Ruiqi Zhong, Samuel R. Bowman, Shi Feng","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-09-19T14:50:34Z","title":"Language Models Learn to Mislead Humans via RLHF"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.12822","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":"sha256:727834885552b1517305fecc940f1a3e4b427966b80587b5d1560b62d37dc236","target":"record","created_at":"2026-07-05T09:45:54Z","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":"40646969e26c3483a945a8a39d02875bab5684919d47deed1671d1fe611b0712","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-09-19T14:50:34Z","title_canon_sha256":"4aec86aaca2d9608eb4014a0a9251f32afbfc7bbb3f7f68b06976e7cf9813d0f"},"schema_version":"1.0","source":{"id":"2409.12822","kind":"arxiv","version":3}},"canonical_sha256":"9ea7f14ab4f70f2d97ecced1fb172b67301e7564e73f410dde1f66147b5a8b03","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"9ea7f14ab4f70f2d97ecced1fb172b67301e7564e73f410dde1f66147b5a8b03","first_computed_at":"2026-07-05T09:45:54.959475Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:45:54.959475Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"3VsPeseANPyZRf8CJ9/CTvApY2Gx+SwVWzqxYnvNyIeupMroEXXB+UXAaS+rJ1Cv6FArPS5GeNN8+npqdRdvAw==","signature_status":"signed_v1","signed_at":"2026-07-05T09:45:54.959943Z","signed_message":"canonical_sha256_bytes"},"source_id":"2409.12822","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:727834885552b1517305fecc940f1a3e4b427966b80587b5d1560b62d37dc236","sha256:6ff0de7a808a0ddd2d4eb497abfe8176133d0223ced02384bebc4fcb7b53b527"],"state_sha256":"4f1c371be434cd43055c619527ba0aba5f69ea9bc52465667a669fcff9093249"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"tDeJDVei5wsSgiSpuCRjUyzTnhTd8a0zh4CGt2p7GJ2BUCn1oF+ToFk65vEUXUZfVo87zgPAw7IDgyzz0PELAQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-13T18:36:14.500026Z","bundle_sha256":"040b97823014df09ce40902c683bed642a8b74ee3e767b3474d3e71cb9f4847f"}}