{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:JD4AUJBFGW2MBILCCXK367VXES","short_pith_number":"pith:JD4AUJBF","schema_version":"1.0","canonical_sha256":"48f80a242535b4c0a16215d5bf7eb72497cca937e0f69f844e7aa3b5035e110b","source":{"kind":"arxiv","id":"2503.01141","version":2},"attestation_state":"computed","paper":{"title":"How Well do LLMs Compress Their Own Chain-of-Thought? A Token Complexity Approach","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ayeong Lee, Ethan Che, Tianyi Peng","submitted_at":"2025-03-03T03:48:20Z","abstract_excerpt":"Chain-of-thought prompting has emerged as a powerful technique for enabling large language models (LLMs) to solve complex reasoning tasks. However, these reasoning chains can be verbose, raising concerns about efficiency. In response, recent works have sought to decrease response lengths through simple prompting strategies (e.g. 'be concise'). In this work, we conduct the first systematic study of the relationship between reasoning length and model performance across a diverse range of compression instructions (e.g. 'use 10 words or less' or 'remove all punctuation'). In doing so, we discover "},"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":"2503.01141","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-03-03T03:48:20Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d35dbc363d5198356219aa80916161dfd15e27fe74ec19f100544d37737454fc","abstract_canon_sha256":"73ef2bd452506d1175f73034bade7f031be06479c7c4ad43224de74d872f0567"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:42:15.350985Z","signature_b64":"u6zuFgtj6kv94fX7UU919uH5P83fPBqmdnlamah8WppCvF+qQIiOvJIZDspGM0XSiMvKGGfA9bhlgrGzckfADg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"48f80a242535b4c0a16215d5bf7eb72497cca937e0f69f844e7aa3b5035e110b","last_reissued_at":"2026-07-05T10:42:15.350525Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:42:15.350525Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Well do LLMs Compress Their Own Chain-of-Thought? A Token Complexity Approach","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ayeong Lee, Ethan Che, Tianyi Peng","submitted_at":"2025-03-03T03:48:20Z","abstract_excerpt":"Chain-of-thought prompting has emerged as a powerful technique for enabling large language models (LLMs) to solve complex reasoning tasks. However, these reasoning chains can be verbose, raising concerns about efficiency. In response, recent works have sought to decrease response lengths through simple prompting strategies (e.g. 'be concise'). In this work, we conduct the first systematic study of the relationship between reasoning length and model performance across a diverse range of compression instructions (e.g. 'use 10 words or less' or 'remove all punctuation'). In doing so, we discover "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.01141","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/2503.01141/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":"2503.01141","created_at":"2026-07-05T10:42:15.350582+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.01141v2","created_at":"2026-07-05T10:42:15.350582+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.01141","created_at":"2026-07-05T10:42:15.350582+00:00"},{"alias_kind":"pith_short_12","alias_value":"JD4AUJBFGW2M","created_at":"2026-07-05T10:42:15.350582+00:00"},{"alias_kind":"pith_short_16","alias_value":"JD4AUJBFGW2MBILC","created_at":"2026-07-05T10:42:15.350582+00:00"},{"alias_kind":"pith_short_8","alias_value":"JD4AUJBF","created_at":"2026-07-05T10:42:15.350582+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08399","citing_title":"Prompt Compression via Activation Aggregation","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2606.03503","citing_title":"ThoughtFold: Folding Reasoning Chains via Introspective Preference Learning","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30832","citing_title":"SLAT: Segment-Level Adaptive Trimming for Efficient CoT Reasoning","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01168","citing_title":"Thinking Economically: A Hierarchical Framework for Adaptive-Complexity Reasoning in LLMs","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2606.17890","citing_title":"Dynamic Rollout Editing for Reducing Overthinking in RL-Trained Reasoning Models","ref_index":23,"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":21,"is_internal_anchor":false},{"citing_arxiv_id":"2503.16419","citing_title":"Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models","ref_index":83,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08776","citing_title":"Reasoning Compression with Mixed-Policy Distillation","ref_index":24,"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":109,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17002","citing_title":"Intelligent Drill-Down: Large Language Model-Driven Drill-Down Technique for Human-AI Collaborative Visual Exploration","ref_index":28,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JD4AUJBFGW2MBILCCXK367VXES","json":"https://pith.science/pith/JD4AUJBFGW2MBILCCXK367VXES.json","graph_json":"https://pith.science/api/pith-number/JD4AUJBFGW2MBILCCXK367VXES/graph.json","events_json":"https://pith.science/api/pith-number/JD4AUJBFGW2MBILCCXK367VXES/events.json","paper":"https://pith.science/paper/JD4AUJBF"},"agent_actions":{"view_html":"https://pith.science/pith/JD4AUJBFGW2MBILCCXK367VXES","download_json":"https://pith.science/pith/JD4AUJBFGW2MBILCCXK367VXES.json","view_paper":"https://pith.science/paper/JD4AUJBF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.01141&json=true","fetch_graph":"https://pith.science/api/pith-number/JD4AUJBFGW2MBILCCXK367VXES/graph.json","fetch_events":"https://pith.science/api/pith-number/JD4AUJBFGW2MBILCCXK367VXES/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JD4AUJBFGW2MBILCCXK367VXES/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JD4AUJBFGW2MBILCCXK367VXES/action/storage_attestation","attest_author":"https://pith.science/pith/JD4AUJBFGW2MBILCCXK367VXES/action/author_attestation","sign_citation":"https://pith.science/pith/JD4AUJBFGW2MBILCCXK367VXES/action/citation_signature","submit_replication":"https://pith.science/pith/JD4AUJBFGW2MBILCCXK367VXES/action/replication_record"}},"created_at":"2026-07-05T10:42:15.350582+00:00","updated_at":"2026-07-05T10:42:15.350582+00:00"}