{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:TDDJRNJJBUI5EV3S5D4J2F6FED","short_pith_number":"pith:TDDJRNJJ","schema_version":"1.0","canonical_sha256":"98c698b5290d11d25772e8f89d17c520ddc574dfbbe954d5f475fb94b54e18b5","source":{"kind":"arxiv","id":"2506.10446","version":1},"attestation_state":"computed","paper":{"title":"Fast on the Easy, Deep on the Hard: Efficient Reasoning via Powered Length Penalty","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Deshu Chen, Hongwei Zhang, Xin Guo, Yifeng Jiao, Yuan Cheng, Zehui Ling","submitted_at":"2025-06-12T07:49:24Z","abstract_excerpt":"Large language models (LLMs) have demonstrated significant advancements in reasoning capabilities, performing well on various challenging benchmarks. Techniques like Chain-of-Thought prompting have been introduced to further improve reasoning. However, these approaches frequently generate longer outputs, which in turn increase computational latency. Although some methods use reinforcement learning to shorten reasoning, they often apply uniform penalties without considering the problem's complexity, leading to suboptimal outcomes. In this study, we seek to enhance the efficiency of LLM reasonin"},"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":"2506.10446","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-12T07:49:24Z","cross_cats_sorted":[],"title_canon_sha256":"178d1a23c05985487d74b54a2d8c7c7673083faf8f2484cf36ac0931ca635e74","abstract_canon_sha256":"301cb76901d446a68768e7598c3eda7d8dfa99331922d4871251bfc89a8660f3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:20:23.882156Z","signature_b64":"8sg18HzPgfU1rXSvkmtjrBmCsJLuyJTHQuvGwa+TLsJr/qBhIBe7BXeIXLnX6hOkHEk0MCTgAhZXmKob+kdIAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"98c698b5290d11d25772e8f89d17c520ddc574dfbbe954d5f475fb94b54e18b5","last_reissued_at":"2026-07-05T11:20:23.881689Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:20:23.881689Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fast on the Easy, Deep on the Hard: Efficient Reasoning via Powered Length Penalty","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Deshu Chen, Hongwei Zhang, Xin Guo, Yifeng Jiao, Yuan Cheng, Zehui Ling","submitted_at":"2025-06-12T07:49:24Z","abstract_excerpt":"Large language models (LLMs) have demonstrated significant advancements in reasoning capabilities, performing well on various challenging benchmarks. Techniques like Chain-of-Thought prompting have been introduced to further improve reasoning. However, these approaches frequently generate longer outputs, which in turn increase computational latency. Although some methods use reinforcement learning to shorten reasoning, they often apply uniform penalties without considering the problem's complexity, leading to suboptimal outcomes. In this study, we seek to enhance the efficiency of LLM reasonin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.10446","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/2506.10446/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":"2506.10446","created_at":"2026-07-05T11:20:23.881745+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.10446v1","created_at":"2026-07-05T11:20:23.881745+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.10446","created_at":"2026-07-05T11:20:23.881745+00:00"},{"alias_kind":"pith_short_12","alias_value":"TDDJRNJJBUI5","created_at":"2026-07-05T11:20:23.881745+00:00"},{"alias_kind":"pith_short_16","alias_value":"TDDJRNJJBUI5EV3S","created_at":"2026-07-05T11:20:23.881745+00:00"},{"alias_kind":"pith_short_8","alias_value":"TDDJRNJJ","created_at":"2026-07-05T11:20:23.881745+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08196","citing_title":"A First-Principles Theory of Slow Thinking and Active Perception","ref_index":103,"is_internal_anchor":true},{"citing_arxiv_id":"2606.22716","citing_title":"Beyond Penalizing Mistakes: Stabilizing Efficiency Training in Large Reasoning Models via Adaptive Correct-Only Rewards","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30832","citing_title":"SLAT: Segment-Level Adaptive Trimming for Efficient CoT Reasoning","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14126","citing_title":"Reinforcement Learning for Tool-Calling Agents in Fast Healthcare Interoperability Resources (FHIR)","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19773","citing_title":"PR-CAD: Progressive Refinement for Unified Controllable and Faithful Text-to-CAD Generation with Large Language Models","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TDDJRNJJBUI5EV3S5D4J2F6FED","json":"https://pith.science/pith/TDDJRNJJBUI5EV3S5D4J2F6FED.json","graph_json":"https://pith.science/api/pith-number/TDDJRNJJBUI5EV3S5D4J2F6FED/graph.json","events_json":"https://pith.science/api/pith-number/TDDJRNJJBUI5EV3S5D4J2F6FED/events.json","paper":"https://pith.science/paper/TDDJRNJJ"},"agent_actions":{"view_html":"https://pith.science/pith/TDDJRNJJBUI5EV3S5D4J2F6FED","download_json":"https://pith.science/pith/TDDJRNJJBUI5EV3S5D4J2F6FED.json","view_paper":"https://pith.science/paper/TDDJRNJJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.10446&json=true","fetch_graph":"https://pith.science/api/pith-number/TDDJRNJJBUI5EV3S5D4J2F6FED/graph.json","fetch_events":"https://pith.science/api/pith-number/TDDJRNJJBUI5EV3S5D4J2F6FED/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TDDJRNJJBUI5EV3S5D4J2F6FED/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TDDJRNJJBUI5EV3S5D4J2F6FED/action/storage_attestation","attest_author":"https://pith.science/pith/TDDJRNJJBUI5EV3S5D4J2F6FED/action/author_attestation","sign_citation":"https://pith.science/pith/TDDJRNJJBUI5EV3S5D4J2F6FED/action/citation_signature","submit_replication":"https://pith.science/pith/TDDJRNJJBUI5EV3S5D4J2F6FED/action/replication_record"}},"created_at":"2026-07-05T11:20:23.881745+00:00","updated_at":"2026-07-05T11:20:23.881745+00:00"}