{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:PAZFEZW5M6HKU7VW2P2QYG2CWX","short_pith_number":"pith:PAZFEZW5","schema_version":"1.0","canonical_sha256":"78325266dd678eaa7eb6d3f50c1b42b5d219d68df5b2487c42b8f43fa8cd709a","source":{"kind":"arxiv","id":"2403.01548","version":3},"attestation_state":"computed","paper":{"title":"In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Junteng Liu, Junxian He, Miao Xiong, Shiqi Chen, Siyang Gao, Teng Xiao, Zhengxuan Wu","submitted_at":"2024-03-03T15:53:41Z","abstract_excerpt":"Large language models (LLMs) frequently hallucinate and produce factual errors, yet our understanding of why they make these errors remains limited. In this study, we delve into the underlying mechanisms of LLM hallucinations from the perspective of inner representations, and discover a salient pattern associated with hallucinations: correct generations tend to have sharper context activations in the hidden states of the in-context tokens, compared to the incorrect ones. Leveraging this insight, we propose an entropy-based metric to quantify the ``sharpness'' among the in-context hidden states"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2403.01548","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-03-03T15:53:41Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"34a0ecb956df3ec5bce810e556709a1ffd472cfdf929a30583f9a78a44432be1","abstract_canon_sha256":"90542017cd0a2cfef449e69aff31331852d078b732d4a773f34d3341d93fba4a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:54:55.196840Z","signature_b64":"COsAAJI6IUnP70E5k/kxGqVrDAQH7Xc2N//k+A7jliK142UBf/k7HVmrMUnAJQLIltgi+qflSHQPhIkLp1+JBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"78325266dd678eaa7eb6d3f50c1b42b5d219d68df5b2487c42b8f43fa8cd709a","last_reissued_at":"2026-07-05T07:54:55.196363Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:54:55.196363Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Junteng Liu, Junxian He, Miao Xiong, Shiqi Chen, Siyang Gao, Teng Xiao, Zhengxuan Wu","submitted_at":"2024-03-03T15:53:41Z","abstract_excerpt":"Large language models (LLMs) frequently hallucinate and produce factual errors, yet our understanding of why they make these errors remains limited. In this study, we delve into the underlying mechanisms of LLM hallucinations from the perspective of inner representations, and discover a salient pattern associated with hallucinations: correct generations tend to have sharper context activations in the hidden states of the in-context tokens, compared to the incorrect ones. Leveraging this insight, we propose an entropy-based metric to quantify the ``sharpness'' among the in-context hidden states"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.01548","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/2403.01548/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2403.01548","created_at":"2026-07-05T07:54:55.196420+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.01548v3","created_at":"2026-07-05T07:54:55.196420+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.01548","created_at":"2026-07-05T07:54:55.196420+00:00"},{"alias_kind":"pith_short_12","alias_value":"PAZFEZW5M6HK","created_at":"2026-07-05T07:54:55.196420+00:00"},{"alias_kind":"pith_short_16","alias_value":"PAZFEZW5M6HKU7VW","created_at":"2026-07-05T07:54:55.196420+00:00"},{"alias_kind":"pith_short_8","alias_value":"PAZFEZW5","created_at":"2026-07-05T07:54:55.196420+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.03022","citing_title":"Hallucinations as Orthogonal Noise: Inference-Time Manifold Alignment via Dynamic Contextual Orthogonalization","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00819","citing_title":"Mitigating Hallucinations in Large Language Models Via Decoder Layer Skipping","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2504.00446","citing_title":"Exposing the Ghost in the Transformer: Abnormal Detection for Large Language Models via Hidden State Forensics","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09492","citing_title":"APCD: Adaptive Path-Contrastive Decoding for Reliable Large Language Model Generation","ref_index":53,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PAZFEZW5M6HKU7VW2P2QYG2CWX","json":"https://pith.science/pith/PAZFEZW5M6HKU7VW2P2QYG2CWX.json","graph_json":"https://pith.science/api/pith-number/PAZFEZW5M6HKU7VW2P2QYG2CWX/graph.json","events_json":"https://pith.science/api/pith-number/PAZFEZW5M6HKU7VW2P2QYG2CWX/events.json","paper":"https://pith.science/paper/PAZFEZW5"},"agent_actions":{"view_html":"https://pith.science/pith/PAZFEZW5M6HKU7VW2P2QYG2CWX","download_json":"https://pith.science/pith/PAZFEZW5M6HKU7VW2P2QYG2CWX.json","view_paper":"https://pith.science/paper/PAZFEZW5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.01548&json=true","fetch_graph":"https://pith.science/api/pith-number/PAZFEZW5M6HKU7VW2P2QYG2CWX/graph.json","fetch_events":"https://pith.science/api/pith-number/PAZFEZW5M6HKU7VW2P2QYG2CWX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PAZFEZW5M6HKU7VW2P2QYG2CWX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PAZFEZW5M6HKU7VW2P2QYG2CWX/action/storage_attestation","attest_author":"https://pith.science/pith/PAZFEZW5M6HKU7VW2P2QYG2CWX/action/author_attestation","sign_citation":"https://pith.science/pith/PAZFEZW5M6HKU7VW2P2QYG2CWX/action/citation_signature","submit_replication":"https://pith.science/pith/PAZFEZW5M6HKU7VW2P2QYG2CWX/action/replication_record"}},"created_at":"2026-07-05T07:54:55.196420+00:00","updated_at":"2026-07-05T07:54:55.196420+00:00"}