{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LYBEBT63THSYABMSQN7TVWXE4S","short_pith_number":"pith:LYBEBT63","schema_version":"1.0","canonical_sha256":"5e0240cfdb99e5800592837f3adae4e4b2fdfea16e7acfcb92cc808978141411","source":{"kind":"arxiv","id":"2504.20168","version":1},"attestation_state":"computed","paper":{"title":"MICE for CATs: Model-Internal Confidence Estimation for Calibrating Agents with Tools","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Benjamin Van Durme, Jason Eisner, Justin Svegliato, Nishant Subramani, Sam Thomson, Yu Su","submitted_at":"2025-04-28T18:06:38Z","abstract_excerpt":"Tool-using agents that act in the world need to be both useful and safe. Well-calibrated model confidences can be used to weigh the risk versus reward of potential actions, but prior work shows that many models are poorly calibrated. Inspired by interpretability literature exploring the internals of models, we propose a novel class of model-internal confidence estimators (MICE) to better assess confidence when calling tools. MICE first decodes from each intermediate layer of the language model using logitLens and then computes similarity scores between each layer's generation and the final out"},"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":"2504.20168","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-28T18:06:38Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"cd5ccd58d52bfd8e3d26528ca685608c8794aadd0a6ced44e4e1bb43f2667a63","abstract_canon_sha256":"af4c8f3e62dea803286b4fa7402017ccebabf68d8d5353570aad1f3017f2ca1a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:55:37.647974Z","signature_b64":"/lqTUvX0dgv/E9gU5ef+RYJiblt8qC1mhxFNq3O9upYe0HBQN5Dv2njMTXi4I1OD8TLFVR082TM+6UPemPvLDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5e0240cfdb99e5800592837f3adae4e4b2fdfea16e7acfcb92cc808978141411","last_reissued_at":"2026-07-05T10:55:37.647572Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:55:37.647572Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MICE for CATs: Model-Internal Confidence Estimation for Calibrating Agents with Tools","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Benjamin Van Durme, Jason Eisner, Justin Svegliato, Nishant Subramani, Sam Thomson, Yu Su","submitted_at":"2025-04-28T18:06:38Z","abstract_excerpt":"Tool-using agents that act in the world need to be both useful and safe. Well-calibrated model confidences can be used to weigh the risk versus reward of potential actions, but prior work shows that many models are poorly calibrated. Inspired by interpretability literature exploring the internals of models, we propose a novel class of model-internal confidence estimators (MICE) to better assess confidence when calling tools. MICE first decodes from each intermediate layer of the language model using logitLens and then computes similarity scores between each layer's generation and the final out"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.20168","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/2504.20168/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":"2504.20168","created_at":"2026-07-05T10:55:37.647626+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.20168v1","created_at":"2026-07-05T10:55:37.647626+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.20168","created_at":"2026-07-05T10:55:37.647626+00:00"},{"alias_kind":"pith_short_12","alias_value":"LYBEBT63THSY","created_at":"2026-07-05T10:55:37.647626+00:00"},{"alias_kind":"pith_short_16","alias_value":"LYBEBT63THSYABMS","created_at":"2026-07-05T10:55:37.647626+00:00"},{"alias_kind":"pith_short_8","alias_value":"LYBEBT63","created_at":"2026-07-05T10:55:37.647626+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LYBEBT63THSYABMSQN7TVWXE4S","json":"https://pith.science/pith/LYBEBT63THSYABMSQN7TVWXE4S.json","graph_json":"https://pith.science/api/pith-number/LYBEBT63THSYABMSQN7TVWXE4S/graph.json","events_json":"https://pith.science/api/pith-number/LYBEBT63THSYABMSQN7TVWXE4S/events.json","paper":"https://pith.science/paper/LYBEBT63"},"agent_actions":{"view_html":"https://pith.science/pith/LYBEBT63THSYABMSQN7TVWXE4S","download_json":"https://pith.science/pith/LYBEBT63THSYABMSQN7TVWXE4S.json","view_paper":"https://pith.science/paper/LYBEBT63","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.20168&json=true","fetch_graph":"https://pith.science/api/pith-number/LYBEBT63THSYABMSQN7TVWXE4S/graph.json","fetch_events":"https://pith.science/api/pith-number/LYBEBT63THSYABMSQN7TVWXE4S/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LYBEBT63THSYABMSQN7TVWXE4S/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LYBEBT63THSYABMSQN7TVWXE4S/action/storage_attestation","attest_author":"https://pith.science/pith/LYBEBT63THSYABMSQN7TVWXE4S/action/author_attestation","sign_citation":"https://pith.science/pith/LYBEBT63THSYABMSQN7TVWXE4S/action/citation_signature","submit_replication":"https://pith.science/pith/LYBEBT63THSYABMSQN7TVWXE4S/action/replication_record"}},"created_at":"2026-07-05T10:55:37.647626+00:00","updated_at":"2026-07-05T10:55:37.647626+00:00"}