{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WTG7LWHIIZ7GJWR6P73M2XJKL2","short_pith_number":"pith:WTG7LWHI","schema_version":"1.0","canonical_sha256":"b4cdf5d8e8467e64da3e7ff6cd5d2a5e9f553333bd490528c8cc615b12597779","source":{"kind":"arxiv","id":"2506.13752","version":1},"attestation_state":"computed","paper":{"title":"Steering LLM Thinking with Budget Guidance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chuang Gan, Junyan Li, Wenshuo Zhao, Yang Zhang","submitted_at":"2025-06-16T17:57:05Z","abstract_excerpt":"Recent deep-thinking large language models often reason extensively to improve performance, but such lengthy reasoning is not always desirable, as it incurs excessive inference costs with disproportionate performance gains. Controlling reasoning length without sacrificing performance is therefore important, but remains challenging, especially under tight thinking budgets. We propose budget guidance, a simple yet effective method for steering the reasoning process of LLMs toward a target budget without requiring any LLM fine-tuning. Our approach introduces a lightweight predictor that models a "},"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":"2506.13752","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-16T17:57:05Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"1bdc41db018936738df2ae5e0bc9c568c370bc66845f23a2a0a94edd1ec20906","abstract_canon_sha256":"2f8ff7355a2bfa54b278914c132de8258e0ce6e00093e461d19a7da13dbd5005"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:22:26.105491Z","signature_b64":"YhplHsS0wei1X8OQ0DW/jOb/g7RAuiFIBwtZ8V9sr2SHeB1w4xRufxG9qn4DwKQ6AgGQRL4xXpFaEY6/4e8QAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b4cdf5d8e8467e64da3e7ff6cd5d2a5e9f553333bd490528c8cc615b12597779","last_reissued_at":"2026-07-05T11:22:26.104948Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:22:26.104948Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Steering LLM Thinking with Budget Guidance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chuang Gan, Junyan Li, Wenshuo Zhao, Yang Zhang","submitted_at":"2025-06-16T17:57:05Z","abstract_excerpt":"Recent deep-thinking large language models often reason extensively to improve performance, but such lengthy reasoning is not always desirable, as it incurs excessive inference costs with disproportionate performance gains. Controlling reasoning length without sacrificing performance is therefore important, but remains challenging, especially under tight thinking budgets. We propose budget guidance, a simple yet effective method for steering the reasoning process of LLMs toward a target budget without requiring any LLM fine-tuning. Our approach introduces a lightweight predictor that models a "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.13752","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/2506.13752/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":"2506.13752","created_at":"2026-07-05T11:22:26.105017+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.13752v1","created_at":"2026-07-05T11:22:26.105017+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.13752","created_at":"2026-07-05T11:22:26.105017+00:00"},{"alias_kind":"pith_short_12","alias_value":"WTG7LWHIIZ7G","created_at":"2026-07-05T11:22:26.105017+00:00"},{"alias_kind":"pith_short_16","alias_value":"WTG7LWHIIZ7GJWR6","created_at":"2026-07-05T11:22:26.105017+00:00"},{"alias_kind":"pith_short_8","alias_value":"WTG7LWHI","created_at":"2026-07-05T11:22:26.105017+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2510.19669","citing_title":"DiffAdapt: Difficulty-Adaptive Reasoning for Token-Efficient LLM Inference","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13647","citing_title":"FlowCompile: An Optimizing Compiler for Structured LLM Workflows","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01111","citing_title":"When Less is Enough: Efficient Inference via Collaborative Reasoning","ref_index":24,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WTG7LWHIIZ7GJWR6P73M2XJKL2","json":"https://pith.science/pith/WTG7LWHIIZ7GJWR6P73M2XJKL2.json","graph_json":"https://pith.science/api/pith-number/WTG7LWHIIZ7GJWR6P73M2XJKL2/graph.json","events_json":"https://pith.science/api/pith-number/WTG7LWHIIZ7GJWR6P73M2XJKL2/events.json","paper":"https://pith.science/paper/WTG7LWHI"},"agent_actions":{"view_html":"https://pith.science/pith/WTG7LWHIIZ7GJWR6P73M2XJKL2","download_json":"https://pith.science/pith/WTG7LWHIIZ7GJWR6P73M2XJKL2.json","view_paper":"https://pith.science/paper/WTG7LWHI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.13752&json=true","fetch_graph":"https://pith.science/api/pith-number/WTG7LWHIIZ7GJWR6P73M2XJKL2/graph.json","fetch_events":"https://pith.science/api/pith-number/WTG7LWHIIZ7GJWR6P73M2XJKL2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WTG7LWHIIZ7GJWR6P73M2XJKL2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WTG7LWHIIZ7GJWR6P73M2XJKL2/action/storage_attestation","attest_author":"https://pith.science/pith/WTG7LWHIIZ7GJWR6P73M2XJKL2/action/author_attestation","sign_citation":"https://pith.science/pith/WTG7LWHIIZ7GJWR6P73M2XJKL2/action/citation_signature","submit_replication":"https://pith.science/pith/WTG7LWHIIZ7GJWR6P73M2XJKL2/action/replication_record"}},"created_at":"2026-07-05T11:22:26.105017+00:00","updated_at":"2026-07-05T11:22:26.105017+00:00"}