{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:C5MQ2RKMEOAM6N3FKBBU6EBXPF","short_pith_number":"pith:C5MQ2RKM","schema_version":"1.0","canonical_sha256":"17590d454c2380cf376550434f103779730b976e7c7ce2806628ffc355c53c72","source":{"kind":"arxiv","id":"2504.06821","version":2},"attestation_state":"computed","paper":{"title":"Inducing Programmatic Skills for Agentic Tasks","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Apurva Gandhi, Daniel Fried, Graham Neubig, Zora Zhiruo Wang","submitted_at":"2025-04-09T12:25:37Z","abstract_excerpt":"To succeed in common digital tasks such as web navigation, agents must carry out a variety of specialized tasks such as searching for products or planning a travel route. To tackle these tasks, agents can bootstrap themselves by learning task-specific skills online through interaction with the web environment. In this work, we demonstrate that programs are an effective representation for skills. We propose agent skill induction (ASI), which allows agents to adapt themselves by inducing, verifying, and utilizing program-based skills on the fly. We start with an evaluation on the WebArena agent "},"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":"2504.06821","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-09T12:25:37Z","cross_cats_sorted":[],"title_canon_sha256":"850b20161d0916382bbcd3471583a6cea8e5ba254492d7d5d03d116daa84666d","abstract_canon_sha256":"0ad726e8a9081f4b3d33ba4733d7d46ef219f2095912b3119f1a74050e9987c5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:01:45.012666Z","signature_b64":"F4vKCCQpFvbaX6sGK2EYb7RbHo79ZFXXmxfwj80Wiv/UdzwAzDzqD2/Pna46NTn1+ZZn8Fmewh+pSPTo72r2Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"17590d454c2380cf376550434f103779730b976e7c7ce2806628ffc355c53c72","last_reissued_at":"2026-07-05T12:01:45.011776Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:01:45.011776Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Inducing Programmatic Skills for Agentic Tasks","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Apurva Gandhi, Daniel Fried, Graham Neubig, Zora Zhiruo Wang","submitted_at":"2025-04-09T12:25:37Z","abstract_excerpt":"To succeed in common digital tasks such as web navigation, agents must carry out a variety of specialized tasks such as searching for products or planning a travel route. To tackle these tasks, agents can bootstrap themselves by learning task-specific skills online through interaction with the web environment. In this work, we demonstrate that programs are an effective representation for skills. We propose agent skill induction (ASI), which allows agents to adapt themselves by inducing, verifying, and utilizing program-based skills on the fly. We start with an evaluation on the WebArena agent "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.06821","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/2504.06821/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":"2504.06821","created_at":"2026-07-05T12:01:45.011834+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.06821v2","created_at":"2026-07-05T12:01:45.011834+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.06821","created_at":"2026-07-05T12:01:45.011834+00:00"},{"alias_kind":"pith_short_12","alias_value":"C5MQ2RKMEOAM","created_at":"2026-07-05T12:01:45.011834+00:00"},{"alias_kind":"pith_short_16","alias_value":"C5MQ2RKMEOAM6N3F","created_at":"2026-07-05T12:01:45.011834+00:00"},{"alias_kind":"pith_short_8","alias_value":"C5MQ2RKM","created_at":"2026-07-05T12:01:45.011834+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":25,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07321","citing_title":"From Atomic Actions to Standard Operating Procedures: Iterative Tool Optimization for Self-Evolving LLM Agents","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2606.26669","citing_title":"SKILL-DISCO: Distilling and Compiling Agent Traces into Reusable Procedural Skills","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18448","citing_title":"VISUALSKILL: Multimodal Skills for Computer-Use Agents","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2606.17645","citing_title":"Beyond Domains: Reusing Web Skills via Transferable Interaction Patterns","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10423","citing_title":"WebChallenger: A Reliable and Efficient Generalist Web Agent","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08755","citing_title":"Co-Evolving Skill Generation and Policy Optimization","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06416","citing_title":"Unsupervised Skill Discovery for Agentic Data Analysis","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12015","citing_title":"SkillSafetyBench: Evaluating Agent Safety under Skill-Facing Attack Surfaces","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07358","citing_title":"A Comprehensive Survey on Agent Skills: Taxonomy, Techniques, and Applications","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29274","citing_title":"Learnable Assessment Skills for LLM-based Automated Scoring: Rubric Construction via Iterative Optimization","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20631","citing_title":"Harnessing Agent Skills: Architectural Patterns and a Reference Architecture for Skill-Mediated LLM Agents","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00510","citing_title":"Skill or Skip? Learning Selective Skill Invocation in Agentic Tasks via Dual-Granularity Preference Learning","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06416","citing_title":"Unsupervised Skill Discovery for Agentic Data Analysis","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12191","citing_title":"Agentic Environment Engineering for Large Language Models: A Survey of Environment Modeling, Synthesis, Evaluation, and Application","ref_index":235,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22177","citing_title":"Maestro: Reinforcement Learning to Orchestrate Hierarchical Model-Skill Ensembles","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07358","citing_title":"A Comprehensive Survey on Agent Skills: Taxonomy, Techniques, and Applications","ref_index":104,"is_internal_anchor":false},{"citing_arxiv_id":"2509.06477","citing_title":"MAS-Bench: A Unified Benchmark for Shortcut-Augmented Hybrid Mobile GUI Agents","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2601.14287","citing_title":"Chain-of-Memory: Lightweight Memory Construction with Dynamic Evolution for LLM Agents","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12015","citing_title":"SkillSafetyBench: Evaluating Agent Safety under Skill-Facing Attack Surfaces","ref_index":84,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11169","citing_title":"OLIVIA: Online Learning via Inference-time Action Adaptation for Decision Making in LLM ReAct Agents","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06811","citing_title":"SkillTrojan: Backdoor Attacks on Skill-Based Agent Systems","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08224","citing_title":"Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering","ref_index":153,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07358","citing_title":"A Comprehensive Survey on Agent Skills: Taxonomy, Techniques, and Applications","ref_index":104,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02913","citing_title":"Generate, Filter, Control, Replay: A Comprehensive Survey of Rollout Strategies for LLM Reinforcement Learning","ref_index":130,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17503","citing_title":"SkillGraph: Self-Evolving Multi-Agent Collaboration with Multimodal Graph Topology","ref_index":38,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/C5MQ2RKMEOAM6N3FKBBU6EBXPF","json":"https://pith.science/pith/C5MQ2RKMEOAM6N3FKBBU6EBXPF.json","graph_json":"https://pith.science/api/pith-number/C5MQ2RKMEOAM6N3FKBBU6EBXPF/graph.json","events_json":"https://pith.science/api/pith-number/C5MQ2RKMEOAM6N3FKBBU6EBXPF/events.json","paper":"https://pith.science/paper/C5MQ2RKM"},"agent_actions":{"view_html":"https://pith.science/pith/C5MQ2RKMEOAM6N3FKBBU6EBXPF","download_json":"https://pith.science/pith/C5MQ2RKMEOAM6N3FKBBU6EBXPF.json","view_paper":"https://pith.science/paper/C5MQ2RKM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.06821&json=true","fetch_graph":"https://pith.science/api/pith-number/C5MQ2RKMEOAM6N3FKBBU6EBXPF/graph.json","fetch_events":"https://pith.science/api/pith-number/C5MQ2RKMEOAM6N3FKBBU6EBXPF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/C5MQ2RKMEOAM6N3FKBBU6EBXPF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/C5MQ2RKMEOAM6N3FKBBU6EBXPF/action/storage_attestation","attest_author":"https://pith.science/pith/C5MQ2RKMEOAM6N3FKBBU6EBXPF/action/author_attestation","sign_citation":"https://pith.science/pith/C5MQ2RKMEOAM6N3FKBBU6EBXPF/action/citation_signature","submit_replication":"https://pith.science/pith/C5MQ2RKMEOAM6N3FKBBU6EBXPF/action/replication_record"}},"created_at":"2026-07-05T12:01:45.011834+00:00","updated_at":"2026-07-05T12:01:45.011834+00:00"}