{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KL2MT7PUPCWZ5AJLMIX53RVHSP","short_pith_number":"pith:KL2MT7PU","schema_version":"1.0","canonical_sha256":"52f4c9fdf478ad9e812b622fddc6a793d49e7a708e2fb2312ea797f84c186da5","source":{"kind":"arxiv","id":"2503.20840","version":2},"attestation_state":"computed","paper":{"title":"CodeTool: Enhancing Programmatic Tool Invocation of LLMs via Process Supervision","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG","cs.SE"],"primary_cat":"cs.AI","authors_text":"Cheng Liu, Fanghua Ye, Feiliang Ren, Haibo Luo, Jian Li, Nan Du, Qiang Gao, Xiaolong Li, Yifei Lu","submitted_at":"2025-03-26T13:19:01Z","abstract_excerpt":"Tool invocation significantly enhances the capabilities of Large Language Models (LLMs), yet challenges persist, particularly in complex task scenarios. Current methods, such as instruction-enhanced reasoning and supervised fine-tuning, often result in unnecessarily long reasoning paths and face difficulties in verifying the correctness of intermediate steps. In this paper, we propose CodeTool, a novel framework for stepwise code generation that improves LLM tool invocation by leveraging the concise and easily verifiable nature of code. CodeTool incorporates two distinct process rewards: the O"},"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":"2503.20840","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-03-26T13:19:01Z","cross_cats_sorted":["cs.CL","cs.LG","cs.SE"],"title_canon_sha256":"5015a5624621d3fd5851cbd1c5eadc4d9117f37278d45de43e68af6acd3dc7ac","abstract_canon_sha256":"03f3701937803e367fc4ffab04a8944127a9b14a8018dfd975928587489f405f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:49:32.909492Z","signature_b64":"QzTD/+q2PXO7TGjpZn0lVlQ1U4vJpbX06B8xzbibcLceP4s2PgRTlAfwGVZ2Rf9dAWlakrYqrMTOQKhhU6ZGBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"52f4c9fdf478ad9e812b622fddc6a793d49e7a708e2fb2312ea797f84c186da5","last_reissued_at":"2026-07-05T11:49:32.908994Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:49:32.908994Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CodeTool: Enhancing Programmatic Tool Invocation of LLMs via Process Supervision","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG","cs.SE"],"primary_cat":"cs.AI","authors_text":"Cheng Liu, Fanghua Ye, Feiliang Ren, Haibo Luo, Jian Li, Nan Du, Qiang Gao, Xiaolong Li, Yifei Lu","submitted_at":"2025-03-26T13:19:01Z","abstract_excerpt":"Tool invocation significantly enhances the capabilities of Large Language Models (LLMs), yet challenges persist, particularly in complex task scenarios. Current methods, such as instruction-enhanced reasoning and supervised fine-tuning, often result in unnecessarily long reasoning paths and face difficulties in verifying the correctness of intermediate steps. In this paper, we propose CodeTool, a novel framework for stepwise code generation that improves LLM tool invocation by leveraging the concise and easily verifiable nature of code. CodeTool incorporates two distinct process rewards: the O"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.20840","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/2503.20840/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":"2503.20840","created_at":"2026-07-05T11:49:32.909053+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.20840v2","created_at":"2026-07-05T11:49:32.909053+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.20840","created_at":"2026-07-05T11:49:32.909053+00:00"},{"alias_kind":"pith_short_12","alias_value":"KL2MT7PUPCWZ","created_at":"2026-07-05T11:49:32.909053+00:00"},{"alias_kind":"pith_short_16","alias_value":"KL2MT7PUPCWZ5AJL","created_at":"2026-07-05T11:49:32.909053+00:00"},{"alias_kind":"pith_short_8","alias_value":"KL2MT7PU","created_at":"2026-07-05T11:49:32.909053+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.17147","citing_title":"CogDual: Enhancing Dual Cognition of LLMs via Reinforcement Learning with Implicit Rule-Based Rewards","ref_index":16,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KL2MT7PUPCWZ5AJLMIX53RVHSP","json":"https://pith.science/pith/KL2MT7PUPCWZ5AJLMIX53RVHSP.json","graph_json":"https://pith.science/api/pith-number/KL2MT7PUPCWZ5AJLMIX53RVHSP/graph.json","events_json":"https://pith.science/api/pith-number/KL2MT7PUPCWZ5AJLMIX53RVHSP/events.json","paper":"https://pith.science/paper/KL2MT7PU"},"agent_actions":{"view_html":"https://pith.science/pith/KL2MT7PUPCWZ5AJLMIX53RVHSP","download_json":"https://pith.science/pith/KL2MT7PUPCWZ5AJLMIX53RVHSP.json","view_paper":"https://pith.science/paper/KL2MT7PU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.20840&json=true","fetch_graph":"https://pith.science/api/pith-number/KL2MT7PUPCWZ5AJLMIX53RVHSP/graph.json","fetch_events":"https://pith.science/api/pith-number/KL2MT7PUPCWZ5AJLMIX53RVHSP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KL2MT7PUPCWZ5AJLMIX53RVHSP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KL2MT7PUPCWZ5AJLMIX53RVHSP/action/storage_attestation","attest_author":"https://pith.science/pith/KL2MT7PUPCWZ5AJLMIX53RVHSP/action/author_attestation","sign_citation":"https://pith.science/pith/KL2MT7PUPCWZ5AJLMIX53RVHSP/action/citation_signature","submit_replication":"https://pith.science/pith/KL2MT7PUPCWZ5AJLMIX53RVHSP/action/replication_record"}},"created_at":"2026-07-05T11:49:32.909053+00:00","updated_at":"2026-07-05T11:49:32.909053+00:00"}