{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BPZE5AAG3Y7RHAYCNZSP4HFIE3","short_pith_number":"pith:BPZE5AAG","schema_version":"1.0","canonical_sha256":"0bf24e8006de3f1383026e64fe1ca826c5a32870d2c2d42ac626b3c5e7580114","source":{"kind":"arxiv","id":"2505.13820","version":5},"attestation_state":"computed","paper":{"title":"Structured Agent Distillation for Large Language Model","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Changdi Yang, Dong Huang, Geng Yuan, Hao Tang, Jun Liu, Peiyan Dong, Pu Zhao, Tianqi Li, Wei Niu, Wenbin Zhang, Xue Lin, Yanzhi Wang, Zhenglun Kong","submitted_at":"2025-05-20T02:01:55Z","abstract_excerpt":"Large language models (LLMs) exhibit strong capabilities as decision-making agents by interleaving reasoning and actions, as seen in ReAct-style frameworks. Yet, their practical deployment is constrained by high inference costs and large model sizes. We propose Structured Agent Distillation, a framework that compresses large LLM-based agents into smaller student models while preserving both reasoning fidelity and action consistency. Unlike standard token-level distillation, our method segments trajectories into [REASON] and [ACT] spans, applying segment-specific losses to align each component "},"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":"2505.13820","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-20T02:01:55Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"62bdfe1f02f13d7a78eb8488699fb6a207bafb4d73028c82ac0ef3d96ca7d645","abstract_canon_sha256":"c76464ae7ab2c3b833040bffb12618b536109c136ea1d4d3469aed86ed75b4ca"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T02:04:40.798106Z","signature_b64":"hbZc50H3SrWF+fxkmwuRjv9pxASpl15+8SIJQUU2MFDnyAGyr/2JGu3SkkotwvvNybmd7C1+5trWLIXrl+iQCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0bf24e8006de3f1383026e64fe1ca826c5a32870d2c2d42ac626b3c5e7580114","last_reissued_at":"2026-05-28T02:04:40.797553Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T02:04:40.797553Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Structured Agent Distillation for Large Language Model","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Changdi Yang, Dong Huang, Geng Yuan, Hao Tang, Jun Liu, Peiyan Dong, Pu Zhao, Tianqi Li, Wei Niu, Wenbin Zhang, Xue Lin, Yanzhi Wang, Zhenglun Kong","submitted_at":"2025-05-20T02:01:55Z","abstract_excerpt":"Large language models (LLMs) exhibit strong capabilities as decision-making agents by interleaving reasoning and actions, as seen in ReAct-style frameworks. Yet, their practical deployment is constrained by high inference costs and large model sizes. We propose Structured Agent Distillation, a framework that compresses large LLM-based agents into smaller student models while preserving both reasoning fidelity and action consistency. Unlike standard token-level distillation, our method segments trajectories into [REASON] and [ACT] spans, applying segment-specific losses to align each component "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.13820","kind":"arxiv","version":5},"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/2505.13820/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":"2505.13820","created_at":"2026-05-28T02:04:40.797629+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.13820v5","created_at":"2026-05-28T02:04:40.797629+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.13820","created_at":"2026-05-28T02:04:40.797629+00:00"},{"alias_kind":"pith_short_12","alias_value":"BPZE5AAG3Y7R","created_at":"2026-05-28T02:04:40.797629+00:00"},{"alias_kind":"pith_short_16","alias_value":"BPZE5AAG3Y7RHAYC","created_at":"2026-05-28T02:04:40.797629+00:00"},{"alias_kind":"pith_short_8","alias_value":"BPZE5AAG","created_at":"2026-05-28T02:04:40.797629+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.07725","citing_title":"SOD: Step-wise On-policy Distillation for Small Language Model Agents","ref_index":16,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BPZE5AAG3Y7RHAYCNZSP4HFIE3","json":"https://pith.science/pith/BPZE5AAG3Y7RHAYCNZSP4HFIE3.json","graph_json":"https://pith.science/api/pith-number/BPZE5AAG3Y7RHAYCNZSP4HFIE3/graph.json","events_json":"https://pith.science/api/pith-number/BPZE5AAG3Y7RHAYCNZSP4HFIE3/events.json","paper":"https://pith.science/paper/BPZE5AAG"},"agent_actions":{"view_html":"https://pith.science/pith/BPZE5AAG3Y7RHAYCNZSP4HFIE3","download_json":"https://pith.science/pith/BPZE5AAG3Y7RHAYCNZSP4HFIE3.json","view_paper":"https://pith.science/paper/BPZE5AAG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.13820&json=true","fetch_graph":"https://pith.science/api/pith-number/BPZE5AAG3Y7RHAYCNZSP4HFIE3/graph.json","fetch_events":"https://pith.science/api/pith-number/BPZE5AAG3Y7RHAYCNZSP4HFIE3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BPZE5AAG3Y7RHAYCNZSP4HFIE3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BPZE5AAG3Y7RHAYCNZSP4HFIE3/action/storage_attestation","attest_author":"https://pith.science/pith/BPZE5AAG3Y7RHAYCNZSP4HFIE3/action/author_attestation","sign_citation":"https://pith.science/pith/BPZE5AAG3Y7RHAYCNZSP4HFIE3/action/citation_signature","submit_replication":"https://pith.science/pith/BPZE5AAG3Y7RHAYCNZSP4HFIE3/action/replication_record"}},"created_at":"2026-05-28T02:04:40.797629+00:00","updated_at":"2026-05-28T02:04:40.797629+00:00"}