{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FVMDBX5NF2STMGQ5BPULCAI4VP","short_pith_number":"pith:FVMDBX5N","schema_version":"1.0","canonical_sha256":"2d5830dfad2ea5361a1d0be8b1011cabd0bee1764adc2d3a5f972d4e005a1506","source":{"kind":"arxiv","id":"2403.08140","version":2},"attestation_state":"computed","paper":{"title":"BAGEL: Bootstrapping Agents by Guiding Exploration with Language","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Christopher Manning, Kenton Lee, Mandar Joshi, Peter Shaw, Shikhar Murty","submitted_at":"2024-03-12T23:59:15Z","abstract_excerpt":"Following natural language instructions by executing actions in digital environments (e.g. web-browsers and REST APIs) is a challenging task for language model (LM) agents. Unfortunately, LM agents often fail to generalize to new environments without human demonstrations. This work presents BAGEL, a method for bootstrapping LM agents without human supervision. BAGEL converts a seed set of randomly explored trajectories or synthetic instructions, into demonstrations, via round-trips between two noisy LM components: an LM labeler which converts a trajectory into a synthetic instruction, and a ze"},"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":"2403.08140","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-03-12T23:59:15Z","cross_cats_sorted":[],"title_canon_sha256":"493faab7ca95db51f8facb315343cd9acda58cac2989fbbcafd3530c6abab891","abstract_canon_sha256":"d9668efb9b8ba9336474514d5c024bd42e9ba4130e75b73ace09a1837bededcc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:29:24.119350Z","signature_b64":"pvnn54j03/4nrM0ZAV6UcAvSS7HkWsaORf3AH+6cqkTqHo1xfsJIZSYdXS/XR9zljf9t4vBYiavhmOg2pEU7CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2d5830dfad2ea5361a1d0be8b1011cabd0bee1764adc2d3a5f972d4e005a1506","last_reissued_at":"2026-07-05T08:29:24.118767Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:29:24.118767Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BAGEL: Bootstrapping Agents by Guiding Exploration with Language","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Christopher Manning, Kenton Lee, Mandar Joshi, Peter Shaw, Shikhar Murty","submitted_at":"2024-03-12T23:59:15Z","abstract_excerpt":"Following natural language instructions by executing actions in digital environments (e.g. web-browsers and REST APIs) is a challenging task for language model (LM) agents. Unfortunately, LM agents often fail to generalize to new environments without human demonstrations. This work presents BAGEL, a method for bootstrapping LM agents without human supervision. BAGEL converts a seed set of randomly explored trajectories or synthetic instructions, into demonstrations, via round-trips between two noisy LM components: an LM labeler which converts a trajectory into a synthetic instruction, and a ze"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.08140","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/2403.08140/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":"2403.08140","created_at":"2026-07-05T08:29:24.118853+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.08140v2","created_at":"2026-07-05T08:29:24.118853+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.08140","created_at":"2026-07-05T08:29:24.118853+00:00"},{"alias_kind":"pith_short_12","alias_value":"FVMDBX5NF2ST","created_at":"2026-07-05T08:29:24.118853+00:00"},{"alias_kind":"pith_short_16","alias_value":"FVMDBX5NF2STMGQ5","created_at":"2026-07-05T08:29:24.118853+00:00"},{"alias_kind":"pith_short_8","alias_value":"FVMDBX5N","created_at":"2026-07-05T08:29:24.118853+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2501.16150","citing_title":"A Comprehensive Survey of Agents for Computer Use: Foundations, Challenges, and Future Directions","ref_index":111,"is_internal_anchor":false},{"citing_arxiv_id":"2409.07429","citing_title":"Agent Workflow Memory","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2405.14573","citing_title":"AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents","ref_index":21,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FVMDBX5NF2STMGQ5BPULCAI4VP","json":"https://pith.science/pith/FVMDBX5NF2STMGQ5BPULCAI4VP.json","graph_json":"https://pith.science/api/pith-number/FVMDBX5NF2STMGQ5BPULCAI4VP/graph.json","events_json":"https://pith.science/api/pith-number/FVMDBX5NF2STMGQ5BPULCAI4VP/events.json","paper":"https://pith.science/paper/FVMDBX5N"},"agent_actions":{"view_html":"https://pith.science/pith/FVMDBX5NF2STMGQ5BPULCAI4VP","download_json":"https://pith.science/pith/FVMDBX5NF2STMGQ5BPULCAI4VP.json","view_paper":"https://pith.science/paper/FVMDBX5N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.08140&json=true","fetch_graph":"https://pith.science/api/pith-number/FVMDBX5NF2STMGQ5BPULCAI4VP/graph.json","fetch_events":"https://pith.science/api/pith-number/FVMDBX5NF2STMGQ5BPULCAI4VP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FVMDBX5NF2STMGQ5BPULCAI4VP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FVMDBX5NF2STMGQ5BPULCAI4VP/action/storage_attestation","attest_author":"https://pith.science/pith/FVMDBX5NF2STMGQ5BPULCAI4VP/action/author_attestation","sign_citation":"https://pith.science/pith/FVMDBX5NF2STMGQ5BPULCAI4VP/action/citation_signature","submit_replication":"https://pith.science/pith/FVMDBX5NF2STMGQ5BPULCAI4VP/action/replication_record"}},"created_at":"2026-07-05T08:29:24.118853+00:00","updated_at":"2026-07-05T08:29:24.118853+00:00"}