{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:NVSUSLHTANHFK7WWHXFLB2M7FV","short_pith_number":"pith:NVSUSLHT","schema_version":"1.0","canonical_sha256":"6d65492cf3034e557ed63dcab0e99f2d57809a5f0778c3758ec3acbd0fac0467","source":{"kind":"arxiv","id":"2408.16090","version":2},"attestation_state":"computed","paper":{"title":"EPO: Hierarchical LLM Agents with Environment Preference Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chen Sun, George Konidaris, Haotian Fu, Qi Zhao","submitted_at":"2024-08-28T18:44:02Z","abstract_excerpt":"Long-horizon decision-making tasks present significant challenges for LLM-based agents due to the need for extensive planning over multiple steps. In this paper, we propose a hierarchical framework that decomposes complex tasks into manageable subgoals, utilizing separate LLMs for subgoal prediction and low-level action generation. To address the challenge of creating training signals for unannotated datasets, we develop a reward model that leverages multimodal environment feedback to automatically generate reward signals. We introduce Environment Preference Optimization (EPO), a novel method "},"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":"2408.16090","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-08-28T18:44:02Z","cross_cats_sorted":[],"title_canon_sha256":"672cb75ceeee59dc3816a96c07da541c47e09c39009eeed555f4ce9f4d98b15f","abstract_canon_sha256":"41feb32f8879bfdd296eb8844c7e22605d587ddfc195bfa171bf5ad5e75cd685"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:15:37.198567Z","signature_b64":"IsUJp/2OkxbCNrAt3f9wRXHBncQ9vkOB9/XALEeGZ7Ahbdd5B89boX9mSCLUUiZagbwQu0OWTaG9XAQ/Ftj4Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6d65492cf3034e557ed63dcab0e99f2d57809a5f0778c3758ec3acbd0fac0467","last_reissued_at":"2026-07-05T09:15:37.198032Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:15:37.198032Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"EPO: Hierarchical LLM Agents with Environment Preference Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chen Sun, George Konidaris, Haotian Fu, Qi Zhao","submitted_at":"2024-08-28T18:44:02Z","abstract_excerpt":"Long-horizon decision-making tasks present significant challenges for LLM-based agents due to the need for extensive planning over multiple steps. In this paper, we propose a hierarchical framework that decomposes complex tasks into manageable subgoals, utilizing separate LLMs for subgoal prediction and low-level action generation. To address the challenge of creating training signals for unannotated datasets, we develop a reward model that leverages multimodal environment feedback to automatically generate reward signals. We introduce Environment Preference Optimization (EPO), a novel method "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.16090","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/2408.16090/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":"2408.16090","created_at":"2026-07-05T09:15:37.198097+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.16090v2","created_at":"2026-07-05T09:15:37.198097+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.16090","created_at":"2026-07-05T09:15:37.198097+00:00"},{"alias_kind":"pith_short_12","alias_value":"NVSUSLHTANHF","created_at":"2026-07-05T09:15:37.198097+00:00"},{"alias_kind":"pith_short_16","alias_value":"NVSUSLHTANHFK7WW","created_at":"2026-07-05T09:15:37.198097+00:00"},{"alias_kind":"pith_short_8","alias_value":"NVSUSLHT","created_at":"2026-07-05T09:15:37.198097+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.14504","citing_title":"When Robots Do the Chores: A Benchmark and Agent for Long-Horizon Household Task Execution","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14504","citing_title":"When Robots Do the Chores: A Benchmark and Agent for Long-Horizon Household Task Execution","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21232","citing_title":"ReCAPA: Hierarchical Predictive Correction to Mitigate Cascading Failures","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21232","citing_title":"ReCAPA: Hierarchical Predictive Correction to Mitigate Cascading Failures","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NVSUSLHTANHFK7WWHXFLB2M7FV","json":"https://pith.science/pith/NVSUSLHTANHFK7WWHXFLB2M7FV.json","graph_json":"https://pith.science/api/pith-number/NVSUSLHTANHFK7WWHXFLB2M7FV/graph.json","events_json":"https://pith.science/api/pith-number/NVSUSLHTANHFK7WWHXFLB2M7FV/events.json","paper":"https://pith.science/paper/NVSUSLHT"},"agent_actions":{"view_html":"https://pith.science/pith/NVSUSLHTANHFK7WWHXFLB2M7FV","download_json":"https://pith.science/pith/NVSUSLHTANHFK7WWHXFLB2M7FV.json","view_paper":"https://pith.science/paper/NVSUSLHT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.16090&json=true","fetch_graph":"https://pith.science/api/pith-number/NVSUSLHTANHFK7WWHXFLB2M7FV/graph.json","fetch_events":"https://pith.science/api/pith-number/NVSUSLHTANHFK7WWHXFLB2M7FV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NVSUSLHTANHFK7WWHXFLB2M7FV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NVSUSLHTANHFK7WWHXFLB2M7FV/action/storage_attestation","attest_author":"https://pith.science/pith/NVSUSLHTANHFK7WWHXFLB2M7FV/action/author_attestation","sign_citation":"https://pith.science/pith/NVSUSLHTANHFK7WWHXFLB2M7FV/action/citation_signature","submit_replication":"https://pith.science/pith/NVSUSLHTANHFK7WWHXFLB2M7FV/action/replication_record"}},"created_at":"2026-07-05T09:15:37.198097+00:00","updated_at":"2026-07-05T09:15:37.198097+00:00"}