{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:624U7R2Q337XFKTDLGLEKDFEUH","short_pith_number":"pith:624U7R2Q","canonical_record":{"source":{"id":"2406.12347","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-18T07:28:15Z","cross_cats_sorted":[],"title_canon_sha256":"326c0036f842afeff5699763de84f94ab4cd892639bc82912877da0c0e5ff95e","abstract_canon_sha256":"45191eda2714c2debd9dbe521e3eaf6723d63262cf66c39177d99c5d8e8461cc"},"schema_version":"1.0"},"canonical_sha256":"f6b94fc750deff72aa635996450ca4a1ea3f501bcb687f4bb81e54614da68f72","source":{"kind":"arxiv","id":"2406.12347","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.12347","created_at":"2026-07-05T08:33:44Z"},{"alias_kind":"arxiv_version","alias_value":"2406.12347v1","created_at":"2026-07-05T08:33:44Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.12347","created_at":"2026-07-05T08:33:44Z"},{"alias_kind":"pith_short_12","alias_value":"624U7R2Q337X","created_at":"2026-07-05T08:33:44Z"},{"alias_kind":"pith_short_16","alias_value":"624U7R2Q337XFKTD","created_at":"2026-07-05T08:33:44Z"},{"alias_kind":"pith_short_8","alias_value":"624U7R2Q","created_at":"2026-07-05T08:33:44Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:624U7R2Q337XFKTDLGLEKDFEUH","target":"record","payload":{"canonical_record":{"source":{"id":"2406.12347","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-18T07:28:15Z","cross_cats_sorted":[],"title_canon_sha256":"326c0036f842afeff5699763de84f94ab4cd892639bc82912877da0c0e5ff95e","abstract_canon_sha256":"45191eda2714c2debd9dbe521e3eaf6723d63262cf66c39177d99c5d8e8461cc"},"schema_version":"1.0"},"canonical_sha256":"f6b94fc750deff72aa635996450ca4a1ea3f501bcb687f4bb81e54614da68f72","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:33:44.710980Z","signature_b64":"BzJrp79MSfMLnIZSSowQQdyWBuivJMcnMoEiIphms643uT1hqYiTS/2qc1/hdLCd3xckJOJqnmP9yDMMt06iAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f6b94fc750deff72aa635996450ca4a1ea3f501bcb687f4bb81e54614da68f72","last_reissued_at":"2026-07-05T08:33:44.710584Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:33:44.710584Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2406.12347","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":"2026-07-05T08:33:44Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AQKyKg7CzSLAItM5lD0tIWvNgyff+g881EpF2OsWHCtn9Y1UCZJ77YfFrJE+7z3fcJ05widBhYSmZAMq9qnzDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T02:27:46.455298Z"},"content_sha256":"670bdfdc3652679a4f7df559d566bd54b6649f70f0793605ff1de51bd35cd86c","schema_version":"1.0","event_id":"sha256:670bdfdc3652679a4f7df559d566bd54b6649f70f0793605ff1de51bd35cd86c"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:624U7R2Q337XFKTDLGLEKDFEUH","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Interpreting Bias in Large Language Models: A Feature-Based Approach","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Lee Ka Wei Roy, Nirmalendu Prakash","submitted_at":"2024-06-18T07:28:15Z","abstract_excerpt":"Large Language Models (LLMs) such as Mistral and LLaMA have showcased remarkable performance across various natural language processing (NLP) tasks. Despite their success, these models inherit social biases from the diverse datasets on which they are trained. This paper investigates the propagation of biases within LLMs through a novel feature-based analytical approach. Drawing inspiration from causal mediation analysis, we hypothesize the evolution of bias-related features and validate them using interpretability techniques like activation and attribution patching. Our contributions are three"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.12347","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/2406.12347/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":"2026-07-05T08:33:44Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XjRYKhplup0MFbfjzVI6WwFfDrgFz2Kj3iBoiswZg7IzdN8i5R8QwZLyy4t+eX2K9DhmHzCRqZZMnEHliKLLBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T02:27:46.455861Z"},"content_sha256":"3fa20fa1c6013528b843aeb99d991c64456abae7f4579ec1ce5461a93e35f8e8","schema_version":"1.0","event_id":"sha256:3fa20fa1c6013528b843aeb99d991c64456abae7f4579ec1ce5461a93e35f8e8"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/624U7R2Q337XFKTDLGLEKDFEUH/bundle.json","state_url":"https://pith.science/pith/624U7R2Q337XFKTDLGLEKDFEUH/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/624U7R2Q337XFKTDLGLEKDFEUH/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":"2026-08-08T02:27:46Z","links":{"resolver":"https://pith.science/pith/624U7R2Q337XFKTDLGLEKDFEUH","bundle":"https://pith.science/pith/624U7R2Q337XFKTDLGLEKDFEUH/bundle.json","state":"https://pith.science/pith/624U7R2Q337XFKTDLGLEKDFEUH/state.json","well_known_bundle":"https://pith.science/.well-known/pith/624U7R2Q337XFKTDLGLEKDFEUH/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:624U7R2Q337XFKTDLGLEKDFEUH","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":"45191eda2714c2debd9dbe521e3eaf6723d63262cf66c39177d99c5d8e8461cc","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-18T07:28:15Z","title_canon_sha256":"326c0036f842afeff5699763de84f94ab4cd892639bc82912877da0c0e5ff95e"},"schema_version":"1.0","source":{"id":"2406.12347","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.12347","created_at":"2026-07-05T08:33:44Z"},{"alias_kind":"arxiv_version","alias_value":"2406.12347v1","created_at":"2026-07-05T08:33:44Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.12347","created_at":"2026-07-05T08:33:44Z"},{"alias_kind":"pith_short_12","alias_value":"624U7R2Q337X","created_at":"2026-07-05T08:33:44Z"},{"alias_kind":"pith_short_16","alias_value":"624U7R2Q337XFKTD","created_at":"2026-07-05T08:33:44Z"},{"alias_kind":"pith_short_8","alias_value":"624U7R2Q","created_at":"2026-07-05T08:33:44Z"}],"graph_snapshots":[{"event_id":"sha256:3fa20fa1c6013528b843aeb99d991c64456abae7f4579ec1ce5461a93e35f8e8","target":"graph","created_at":"2026-07-05T08:33:44Z","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/2406.12347/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large Language Models (LLMs) such as Mistral and LLaMA have showcased remarkable performance across various natural language processing (NLP) tasks. Despite their success, these models inherit social biases from the diverse datasets on which they are trained. This paper investigates the propagation of biases within LLMs through a novel feature-based analytical approach. Drawing inspiration from causal mediation analysis, we hypothesize the evolution of bias-related features and validate them using interpretability techniques like activation and attribution patching. Our contributions are three","authors_text":"Lee Ka Wei Roy, Nirmalendu Prakash","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-18T07:28:15Z","title":"Interpreting Bias in Large Language Models: A Feature-Based Approach"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.12347","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":"sha256:670bdfdc3652679a4f7df559d566bd54b6649f70f0793605ff1de51bd35cd86c","target":"record","created_at":"2026-07-05T08:33:44Z","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":"45191eda2714c2debd9dbe521e3eaf6723d63262cf66c39177d99c5d8e8461cc","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-18T07:28:15Z","title_canon_sha256":"326c0036f842afeff5699763de84f94ab4cd892639bc82912877da0c0e5ff95e"},"schema_version":"1.0","source":{"id":"2406.12347","kind":"arxiv","version":1}},"canonical_sha256":"f6b94fc750deff72aa635996450ca4a1ea3f501bcb687f4bb81e54614da68f72","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"f6b94fc750deff72aa635996450ca4a1ea3f501bcb687f4bb81e54614da68f72","first_computed_at":"2026-07-05T08:33:44.710584Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:33:44.710584Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"BzJrp79MSfMLnIZSSowQQdyWBuivJMcnMoEiIphms643uT1hqYiTS/2qc1/hdLCd3xckJOJqnmP9yDMMt06iAw==","signature_status":"signed_v1","signed_at":"2026-07-05T08:33:44.710980Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.12347","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:670bdfdc3652679a4f7df559d566bd54b6649f70f0793605ff1de51bd35cd86c","sha256:3fa20fa1c6013528b843aeb99d991c64456abae7f4579ec1ce5461a93e35f8e8"],"state_sha256":"b48ea2483d0cd8c32372753630dc96ce88094a1e22cc78e551aa7639191f6b21"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"mfySuLkfbTP7C+XopAfBqeTLK4qLEpqmkpOeTQ9FL8xSFg7MWJoEIAFc9oQkfWDnOLuinffh6pIwy2EOzSu/Bw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-08T02:27:46.460349Z","bundle_sha256":"b20e14ae4f751fe44d6bcf826bb13872c5f712d9b59acbdfedf287212ae0daf8"}}