{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RRLPFGP47SVNNJIVY75VAYT4JO","short_pith_number":"pith:RRLPFGP4","schema_version":"1.0","canonical_sha256":"8c56f299fcfcaad6a515c7fb50627c4bab772a2a21cf012a00d2278ebca7ee67","source":{"kind":"arxiv","id":"2603.05488","version":4},"attestation_state":"computed","paper":{"title":"Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thought","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Annabel Ma, Atticus Geiger, Eric Bigelow, Jack Merullo, Max Loeffler, Owen Lewis, Raphael Sarfati, Siddharth Boppana","submitted_at":"2026-03-05T18:55:16Z","abstract_excerpt":"We provide evidence of performative chain-of-thought (CoT) in reasoning models, where a model becomes strongly confident in its final answer, but continues generating tokens without revealing its internal belief. Our analysis compares activation probing, early forced answering, and a CoT monitor across two large models (DeepSeek-R1 671B & GPT-OSS 120B) and find task difficulty-specific differences: The model's final answer is decodable from activations far earlier in CoT than a monitor is able to say, especially for easy recall-based MMLU questions. We contrast this with genuine reasoning in d"},"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":"2603.05488","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-05T18:55:16Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"ae865c7223b341ba7b77aa1d2f7188fd48432615cce3b932ae71df3a370a05b9","abstract_canon_sha256":"9c86826ca93063a56e5b4aecb6910dde23b1d6d567a35cbda8ada5a0e620483a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T02:05:42.544170Z","signature_b64":"naKCHR4zuxSOS70zt2fGkH8O10TokJjd1HQ0fPyHahl6rHErHf2B88x4E8/rQ9owcSM3a+IHF80pfY+Qx3p6Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8c56f299fcfcaad6a515c7fb50627c4bab772a2a21cf012a00d2278ebca7ee67","last_reissued_at":"2026-05-29T02:05:42.543482Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T02:05:42.543482Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thought","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Annabel Ma, Atticus Geiger, Eric Bigelow, Jack Merullo, Max Loeffler, Owen Lewis, Raphael Sarfati, Siddharth Boppana","submitted_at":"2026-03-05T18:55:16Z","abstract_excerpt":"We provide evidence of performative chain-of-thought (CoT) in reasoning models, where a model becomes strongly confident in its final answer, but continues generating tokens without revealing its internal belief. Our analysis compares activation probing, early forced answering, and a CoT monitor across two large models (DeepSeek-R1 671B & GPT-OSS 120B) and find task difficulty-specific differences: The model's final answer is decodable from activations far earlier in CoT than a monitor is able to say, especially for easy recall-based MMLU questions. We contrast this with genuine reasoning in d"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.05488","kind":"arxiv","version":4},"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/2603.05488/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":"2603.05488","created_at":"2026-05-29T02:05:42.543579+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.05488v4","created_at":"2026-05-29T02:05:42.543579+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.05488","created_at":"2026-05-29T02:05:42.543579+00:00"},{"alias_kind":"pith_short_12","alias_value":"RRLPFGP47SVN","created_at":"2026-05-29T02:05:42.543579+00:00"},{"alias_kind":"pith_short_16","alias_value":"RRLPFGP47SVNNJIV","created_at":"2026-05-29T02:05:42.543579+00:00"},{"alias_kind":"pith_short_8","alias_value":"RRLPFGP4","created_at":"2026-05-29T02:05:42.543579+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":8,"sample":[{"citing_arxiv_id":"2605.20745","citing_title":"The Hidden Signal of Verifier Strictness: Controlling and Improving Step-Wise Verification via Selective Latent Steering","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20075","citing_title":"CopT: Contrastive On-Policy Thinking with Continuous Spaces for General and Agentic Reasoning","ref_index":5,"is_internal_anchor":true},{"citing_arxiv_id":"2604.02371","citing_title":"Internalized Reasoning for Long-Context Visual Document Understanding","ref_index":4,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11746","citing_title":"When Reasoning Traces Become Performative: Step-Level Evidence that Chain-of-Thought Is an Imperfect Oversight Channel","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12412","citing_title":"Stories in Space: In-Context Learning Trajectories in Conceptual Belief Space","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2604.15726","citing_title":"LLM Reasoning Is Latent, Not the Chain of Thought","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2604.16913","citing_title":"The Cognitive Penalty: Ablating System 1 and System 2 Reasoning in Edge-Native SLMs for Decentralized Consensus","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05851","citing_title":"Hypothesis generation and updating in large language models","ref_index":14,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RRLPFGP47SVNNJIVY75VAYT4JO","json":"https://pith.science/pith/RRLPFGP47SVNNJIVY75VAYT4JO.json","graph_json":"https://pith.science/api/pith-number/RRLPFGP47SVNNJIVY75VAYT4JO/graph.json","events_json":"https://pith.science/api/pith-number/RRLPFGP47SVNNJIVY75VAYT4JO/events.json","paper":"https://pith.science/paper/RRLPFGP4"},"agent_actions":{"view_html":"https://pith.science/pith/RRLPFGP47SVNNJIVY75VAYT4JO","download_json":"https://pith.science/pith/RRLPFGP47SVNNJIVY75VAYT4JO.json","view_paper":"https://pith.science/paper/RRLPFGP4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.05488&json=true","fetch_graph":"https://pith.science/api/pith-number/RRLPFGP47SVNNJIVY75VAYT4JO/graph.json","fetch_events":"https://pith.science/api/pith-number/RRLPFGP47SVNNJIVY75VAYT4JO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RRLPFGP47SVNNJIVY75VAYT4JO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RRLPFGP47SVNNJIVY75VAYT4JO/action/storage_attestation","attest_author":"https://pith.science/pith/RRLPFGP47SVNNJIVY75VAYT4JO/action/author_attestation","sign_citation":"https://pith.science/pith/RRLPFGP47SVNNJIVY75VAYT4JO/action/citation_signature","submit_replication":"https://pith.science/pith/RRLPFGP47SVNNJIVY75VAYT4JO/action/replication_record"}},"created_at":"2026-05-29T02:05:42.543579+00:00","updated_at":"2026-05-29T02:05:42.543579+00:00"}