{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:Q4VAYN5G2GCTH4LBWI2MABMG3Q","short_pith_number":"pith:Q4VAYN5G","schema_version":"1.0","canonical_sha256":"872a0c37a6d18533f161b234c00586dc04cbba86486b8609c05b8a7318bb2b93","source":{"kind":"arxiv","id":"2507.16488","version":1},"attestation_state":"computed","paper":{"title":"ICR Probe: Tracking Hidden State Dynamics for Reliable Hallucination Detection in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Huixuan Zhang, Junzhe Zhang, Xiaojun Wan, Xinyu Hu, Zhenliang Zhang","submitted_at":"2025-07-22T11:44:26Z","abstract_excerpt":"Large language models (LLMs) excel at various natural language processing tasks, but their tendency to generate hallucinations undermines their reliability. Existing hallucination detection methods leveraging hidden states predominantly focus on static and isolated representations, overlooking their dynamic evolution across layers, which limits efficacy. To address this limitation, we shift the focus to the hidden state update process and introduce a novel metric, the ICR Score (Information Contribution to Residual Stream), which quantifies the contribution of modules to the hidden states' upd"},"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":"2507.16488","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-07-22T11:44:26Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"88ba1d0a1c727a7f4af41f72fc0382eb2a9c4dd2922e8cab5454a6d9728a0680","abstract_canon_sha256":"f876624d366f5d63e75ccc9a74af210d068dbb825c59b4535b48613af43c0c9d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:41:10.373013Z","signature_b64":"ffEzirfzLYZB3zSKfIhmevL1mU9E8MPSTG7lqSINmeRjWsagpUD+p2TkqCgXPGvhB2XDDXdvgbUsR9wvjo+eDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"872a0c37a6d18533f161b234c00586dc04cbba86486b8609c05b8a7318bb2b93","last_reissued_at":"2026-07-05T11:41:10.372593Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:41:10.372593Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ICR Probe: Tracking Hidden State Dynamics for Reliable Hallucination Detection in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Huixuan Zhang, Junzhe Zhang, Xiaojun Wan, Xinyu Hu, Zhenliang Zhang","submitted_at":"2025-07-22T11:44:26Z","abstract_excerpt":"Large language models (LLMs) excel at various natural language processing tasks, but their tendency to generate hallucinations undermines their reliability. Existing hallucination detection methods leveraging hidden states predominantly focus on static and isolated representations, overlooking their dynamic evolution across layers, which limits efficacy. To address this limitation, we shift the focus to the hidden state update process and introduce a novel metric, the ICR Score (Information Contribution to Residual Stream), which quantifies the contribution of modules to the hidden states' upd"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.16488","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/2507.16488/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":"2507.16488","created_at":"2026-07-05T11:41:10.372645+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.16488v1","created_at":"2026-07-05T11:41:10.372645+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.16488","created_at":"2026-07-05T11:41:10.372645+00:00"},{"alias_kind":"pith_short_12","alias_value":"Q4VAYN5G2GCT","created_at":"2026-07-05T11:41:10.372645+00:00"},{"alias_kind":"pith_short_16","alias_value":"Q4VAYN5G2GCTH4LB","created_at":"2026-07-05T11:41:10.372645+00:00"},{"alias_kind":"pith_short_8","alias_value":"Q4VAYN5G","created_at":"2026-07-05T11:41:10.372645+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22864","citing_title":"When AUC 0.998 Is Not Enough: A Candidate Evaluation Protocol for Hidden-State Probes of Indirect Prompt Injection in Multimodal Computer-Use Agents","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15945","citing_title":"RAGognizer: Hallucination-Aware Fine-Tuning via Detection Head Integration","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Q4VAYN5G2GCTH4LBWI2MABMG3Q","json":"https://pith.science/pith/Q4VAYN5G2GCTH4LBWI2MABMG3Q.json","graph_json":"https://pith.science/api/pith-number/Q4VAYN5G2GCTH4LBWI2MABMG3Q/graph.json","events_json":"https://pith.science/api/pith-number/Q4VAYN5G2GCTH4LBWI2MABMG3Q/events.json","paper":"https://pith.science/paper/Q4VAYN5G"},"agent_actions":{"view_html":"https://pith.science/pith/Q4VAYN5G2GCTH4LBWI2MABMG3Q","download_json":"https://pith.science/pith/Q4VAYN5G2GCTH4LBWI2MABMG3Q.json","view_paper":"https://pith.science/paper/Q4VAYN5G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.16488&json=true","fetch_graph":"https://pith.science/api/pith-number/Q4VAYN5G2GCTH4LBWI2MABMG3Q/graph.json","fetch_events":"https://pith.science/api/pith-number/Q4VAYN5G2GCTH4LBWI2MABMG3Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Q4VAYN5G2GCTH4LBWI2MABMG3Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Q4VAYN5G2GCTH4LBWI2MABMG3Q/action/storage_attestation","attest_author":"https://pith.science/pith/Q4VAYN5G2GCTH4LBWI2MABMG3Q/action/author_attestation","sign_citation":"https://pith.science/pith/Q4VAYN5G2GCTH4LBWI2MABMG3Q/action/citation_signature","submit_replication":"https://pith.science/pith/Q4VAYN5G2GCTH4LBWI2MABMG3Q/action/replication_record"}},"created_at":"2026-07-05T11:41:10.372645+00:00","updated_at":"2026-07-05T11:41:10.372645+00:00"}