{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EUW2RXX4XOEPIXAHYPYH4VJLNZ","short_pith_number":"pith:EUW2RXX4","schema_version":"1.0","canonical_sha256":"252da8defcbb88f45c07c3f07e552b6e6e4eb278a729ee69c90fd9c36c7b1a1b","source":{"kind":"arxiv","id":"2606.00206","version":1},"attestation_state":"computed","paper":{"title":"Quantized Reasoning Models Think They Need to Think Longer, but They Do Not","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Polina Kirichenko, Sanae Lotfi, Steven Li, Zechun Liu","submitted_at":"2026-05-29T18:00:00Z","abstract_excerpt":"Post-training quantization (PTQ) is widely used to deploy large language models efficiently, but its effect on reasoning models is not well understood. Across math, coding, and science QA, we find that aggressive PTQ reduces accuracy while increasing chain-of-thought (CoT) length. Surprisingly, we show that in up to 52% of the quantized models' failures, models reach the right answer in intermediate reasoning steps but do not output it as a final answer. To understand why quantization leads to this increase in overthinking errors, we measure the token-level KL divergence between quantized and "},"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":"2606.00206","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-29T18:00:00Z","cross_cats_sorted":[],"title_canon_sha256":"d4edd90b5e4478e88d05033e2a2013d8be378c40e3a1379aa22fc3dda14d3bc9","abstract_canon_sha256":"ba8ffd84e7b7414bc4d72dc35eef0e8e0f69a713a54a3be1bdc5656ed1ffc1ce"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:03:21.526377Z","signature_b64":"bCgENEpjsmMGjmqIKXGqQ+cZrJHzN/1OQpyWnJWja6AmK4Uqz4To8h0SCvPYnQxbNcSSlOCyVhhPGaEy7JhPAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"252da8defcbb88f45c07c3f07e552b6e6e4eb278a729ee69c90fd9c36c7b1a1b","last_reissued_at":"2026-06-02T01:03:21.526038Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:03:21.526038Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Quantized Reasoning Models Think They Need to Think Longer, but They Do Not","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Polina Kirichenko, Sanae Lotfi, Steven Li, Zechun Liu","submitted_at":"2026-05-29T18:00:00Z","abstract_excerpt":"Post-training quantization (PTQ) is widely used to deploy large language models efficiently, but its effect on reasoning models is not well understood. Across math, coding, and science QA, we find that aggressive PTQ reduces accuracy while increasing chain-of-thought (CoT) length. Surprisingly, we show that in up to 52% of the quantized models' failures, models reach the right answer in intermediate reasoning steps but do not output it as a final answer. To understand why quantization leads to this increase in overthinking errors, we measure the token-level KL divergence between quantized and "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.00206","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/2606.00206/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":"2606.00206","created_at":"2026-06-02T01:03:21.526092+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.00206v1","created_at":"2026-06-02T01:03:21.526092+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.00206","created_at":"2026-06-02T01:03:21.526092+00:00"},{"alias_kind":"pith_short_12","alias_value":"EUW2RXX4XOEP","created_at":"2026-06-02T01:03:21.526092+00:00"},{"alias_kind":"pith_short_16","alias_value":"EUW2RXX4XOEPIXAH","created_at":"2026-06-02T01:03:21.526092+00:00"},{"alias_kind":"pith_short_8","alias_value":"EUW2RXX4","created_at":"2026-06-02T01:03:21.526092+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/EUW2RXX4XOEPIXAHYPYH4VJLNZ","json":"https://pith.science/pith/EUW2RXX4XOEPIXAHYPYH4VJLNZ.json","graph_json":"https://pith.science/api/pith-number/EUW2RXX4XOEPIXAHYPYH4VJLNZ/graph.json","events_json":"https://pith.science/api/pith-number/EUW2RXX4XOEPIXAHYPYH4VJLNZ/events.json","paper":"https://pith.science/paper/EUW2RXX4"},"agent_actions":{"view_html":"https://pith.science/pith/EUW2RXX4XOEPIXAHYPYH4VJLNZ","download_json":"https://pith.science/pith/EUW2RXX4XOEPIXAHYPYH4VJLNZ.json","view_paper":"https://pith.science/paper/EUW2RXX4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.00206&json=true","fetch_graph":"https://pith.science/api/pith-number/EUW2RXX4XOEPIXAHYPYH4VJLNZ/graph.json","fetch_events":"https://pith.science/api/pith-number/EUW2RXX4XOEPIXAHYPYH4VJLNZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EUW2RXX4XOEPIXAHYPYH4VJLNZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EUW2RXX4XOEPIXAHYPYH4VJLNZ/action/storage_attestation","attest_author":"https://pith.science/pith/EUW2RXX4XOEPIXAHYPYH4VJLNZ/action/author_attestation","sign_citation":"https://pith.science/pith/EUW2RXX4XOEPIXAHYPYH4VJLNZ/action/citation_signature","submit_replication":"https://pith.science/pith/EUW2RXX4XOEPIXAHYPYH4VJLNZ/action/replication_record"}},"created_at":"2026-06-02T01:03:21.526092+00:00","updated_at":"2026-06-02T01:03:21.526092+00:00"}