{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5TKPLU3JVHKAHHO5XCVV7QOFJG","short_pith_number":"pith:5TKPLU3J","schema_version":"1.0","canonical_sha256":"ecd4f5d369a9d4039dddb8ab5fc1c549a4b456871c0d88578c9eafd88d556e2c","source":{"kind":"arxiv","id":"2504.06514","version":2},"attestation_state":"computed","paper":{"title":"Missing Premise exacerbates Overthinking: Are Reasoning Models losing Critical Thinking Skill?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Chenrui Fan, Lichao Sun, Ming Li, Tianyi Zhou","submitted_at":"2025-04-09T01:25:27Z","abstract_excerpt":"We find that the response length of reasoning LLMs, whether trained by reinforcement learning or supervised learning, drastically increases for ill-posed questions with missing premises (MiP), ending up with redundant and ineffective thinking. This newly introduced scenario exacerbates the general overthinking issue to a large extent, which we name as the MiP-Overthinking. Such failures are against the ``test-time scaling law'' but have been widely observed on multiple datasets we curated with MiP, indicating the harm of cheap overthinking and a lack of critical thinking. Surprisingly, LLMs no"},"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":"2504.06514","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-04-09T01:25:27Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"80aefafb5301c4821763c57657505efe45e23c6f593e440253044b1093dcdf3e","abstract_canon_sha256":"ea1eea78ea0e9252c34617bce4437e8e6304dd1a7e5a5a4585900fdd7f8de0d6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:47:39.193409Z","signature_b64":"cURpzovWay1nVgUtV+zrxlHWqSNha1+SvJ/X+0zCg1mqelNfjAYBjmNtQ8nMO55qwx4QFJy/cL9bs5CsouRrCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ecd4f5d369a9d4039dddb8ab5fc1c549a4b456871c0d88578c9eafd88d556e2c","last_reissued_at":"2026-07-05T10:47:39.192925Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:47:39.192925Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Missing Premise exacerbates Overthinking: Are Reasoning Models losing Critical Thinking Skill?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Chenrui Fan, Lichao Sun, Ming Li, Tianyi Zhou","submitted_at":"2025-04-09T01:25:27Z","abstract_excerpt":"We find that the response length of reasoning LLMs, whether trained by reinforcement learning or supervised learning, drastically increases for ill-posed questions with missing premises (MiP), ending up with redundant and ineffective thinking. This newly introduced scenario exacerbates the general overthinking issue to a large extent, which we name as the MiP-Overthinking. Such failures are against the ``test-time scaling law'' but have been widely observed on multiple datasets we curated with MiP, indicating the harm of cheap overthinking and a lack of critical thinking. Surprisingly, LLMs no"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.06514","kind":"arxiv","version":2},"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/2504.06514/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":"2504.06514","created_at":"2026-07-05T10:47:39.192983+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.06514v2","created_at":"2026-07-05T10:47:39.192983+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.06514","created_at":"2026-07-05T10:47:39.192983+00:00"},{"alias_kind":"pith_short_12","alias_value":"5TKPLU3JVHKA","created_at":"2026-07-05T10:47:39.192983+00:00"},{"alias_kind":"pith_short_16","alias_value":"5TKPLU3JVHKAHHO5","created_at":"2026-07-05T10:47:39.192983+00:00"},{"alias_kind":"pith_short_8","alias_value":"5TKPLU3J","created_at":"2026-07-05T10:47:39.192983+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":13,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18709","citing_title":"LLMs Struggle to Measure What Distinguishes Students of Different Proficiency Levels: A Study of Item Discrimination in Reading Comprehension Assessment","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00206","citing_title":"Quantized Reasoning Models Think They Need to Think Longer, but They Do Not","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28186","citing_title":"Cognitive Episodes in LLM Reasoning Traces Enable Interpretable Human Item Difficulty Prediction","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28070","citing_title":"Bridging the Detection-to-Abstention Gap in Reasoning Models under Insufficient Information","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2602.08324","citing_title":"Towards Efficient Large Language Reasoning Models via Extreme-Ratio Chain-of-Thought Compression","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2511.22396","citing_title":"Asking like Socrates: Socrates helps VLMs understand remote sensing images","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2512.19995","citing_title":"Schoenfeld's Anatomy of Mathematical Reasoning by Language Models","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13338","citing_title":"Inducing Overthink: Hierarchical Genetic Algorithm-based DoS Attack on Black-Box Large Language Reasoning Models","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13338","citing_title":"Inducing Overthink: Hierarchical Genetic Algorithm-based DoS Attack on Black-Box Large Language Reasoning Models","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2503.16419","citing_title":"Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06165","citing_title":"Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost","ref_index":201,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07316","citing_title":"Implicit Compression Regularization: Concise Reasoning via Internal Shorter Distributions in RL Post-Training","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19656","citing_title":"Pause or Fabricate? Training Language Models for Grounded Reasoning","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5TKPLU3JVHKAHHO5XCVV7QOFJG","json":"https://pith.science/pith/5TKPLU3JVHKAHHO5XCVV7QOFJG.json","graph_json":"https://pith.science/api/pith-number/5TKPLU3JVHKAHHO5XCVV7QOFJG/graph.json","events_json":"https://pith.science/api/pith-number/5TKPLU3JVHKAHHO5XCVV7QOFJG/events.json","paper":"https://pith.science/paper/5TKPLU3J"},"agent_actions":{"view_html":"https://pith.science/pith/5TKPLU3JVHKAHHO5XCVV7QOFJG","download_json":"https://pith.science/pith/5TKPLU3JVHKAHHO5XCVV7QOFJG.json","view_paper":"https://pith.science/paper/5TKPLU3J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.06514&json=true","fetch_graph":"https://pith.science/api/pith-number/5TKPLU3JVHKAHHO5XCVV7QOFJG/graph.json","fetch_events":"https://pith.science/api/pith-number/5TKPLU3JVHKAHHO5XCVV7QOFJG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5TKPLU3JVHKAHHO5XCVV7QOFJG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5TKPLU3JVHKAHHO5XCVV7QOFJG/action/storage_attestation","attest_author":"https://pith.science/pith/5TKPLU3JVHKAHHO5XCVV7QOFJG/action/author_attestation","sign_citation":"https://pith.science/pith/5TKPLU3JVHKAHHO5XCVV7QOFJG/action/citation_signature","submit_replication":"https://pith.science/pith/5TKPLU3JVHKAHHO5XCVV7QOFJG/action/replication_record"}},"created_at":"2026-07-05T10:47:39.192983+00:00","updated_at":"2026-07-05T10:47:39.192983+00:00"}