{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:JZFBZH74GSE6SSYIU6LZULQUSC","short_pith_number":"pith:JZFBZH74","schema_version":"1.0","canonical_sha256":"4e4a1c9ffc3489e94b08a7979a2e1490abae3af62b6a3edbb9f3fdd1a0224012","source":{"kind":"arxiv","id":"2406.00645","version":2},"attestation_state":"computed","paper":{"title":"FuRL: Visual-Language Models as Fuzzy Rewards for Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Benoit Boulet, Di Wu, Haichao Zhang, Wei Xu, Yuwei Fu","submitted_at":"2024-06-02T07:20:08Z","abstract_excerpt":"In this work, we investigate how to leverage pre-trained visual-language models (VLM) for online Reinforcement Learning (RL). In particular, we focus on sparse reward tasks with pre-defined textual task descriptions. We first identify the problem of reward misalignment when applying VLM as a reward in RL tasks. To address this issue, we introduce a lightweight fine-tuning method, named Fuzzy VLM reward-aided RL (FuRL), based on reward alignment and relay RL. Specifically, we enhance the performance of SAC/DrQ baseline agents on sparse reward tasks by fine-tuning VLM representations and using r"},"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":"2406.00645","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-02T07:20:08Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"d005134e17128808864ea15dbd7c5e7451290624bd12c0fa5f3c6c891f644c3f","abstract_canon_sha256":"193e2262951716436dce8f4884352a05296a285a6392d9106b018e38e0a2a975"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:27:26.521839Z","signature_b64":"+shRwonwjp8mew6rUappqo58QeLOrcEN5BdKGwuC58DTtmB/7E4/j9Qfhih72cHNmJiwVEaOV6s7QllO/aBLAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4e4a1c9ffc3489e94b08a7979a2e1490abae3af62b6a3edbb9f3fdd1a0224012","last_reissued_at":"2026-07-05T08:27:26.521422Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:27:26.521422Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FuRL: Visual-Language Models as Fuzzy Rewards for Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Benoit Boulet, Di Wu, Haichao Zhang, Wei Xu, Yuwei Fu","submitted_at":"2024-06-02T07:20:08Z","abstract_excerpt":"In this work, we investigate how to leverage pre-trained visual-language models (VLM) for online Reinforcement Learning (RL). In particular, we focus on sparse reward tasks with pre-defined textual task descriptions. We first identify the problem of reward misalignment when applying VLM as a reward in RL tasks. To address this issue, we introduce a lightweight fine-tuning method, named Fuzzy VLM reward-aided RL (FuRL), based on reward alignment and relay RL. Specifically, we enhance the performance of SAC/DrQ baseline agents on sparse reward tasks by fine-tuning VLM representations and using r"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.00645","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/2406.00645/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":"2406.00645","created_at":"2026-07-05T08:27:26.521479+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.00645v2","created_at":"2026-07-05T08:27:26.521479+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.00645","created_at":"2026-07-05T08:27:26.521479+00:00"},{"alias_kind":"pith_short_12","alias_value":"JZFBZH74GSE6","created_at":"2026-07-05T08:27:26.521479+00:00"},{"alias_kind":"pith_short_16","alias_value":"JZFBZH74GSE6SSYI","created_at":"2026-07-05T08:27:26.521479+00:00"},{"alias_kind":"pith_short_8","alias_value":"JZFBZH74","created_at":"2026-07-05T08:27:26.521479+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.01616","citing_title":"Preference VLM: Leveraging VLMs for Scalable Preference-Based Reinforcement Learning","ref_index":2016,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JZFBZH74GSE6SSYIU6LZULQUSC","json":"https://pith.science/pith/JZFBZH74GSE6SSYIU6LZULQUSC.json","graph_json":"https://pith.science/api/pith-number/JZFBZH74GSE6SSYIU6LZULQUSC/graph.json","events_json":"https://pith.science/api/pith-number/JZFBZH74GSE6SSYIU6LZULQUSC/events.json","paper":"https://pith.science/paper/JZFBZH74"},"agent_actions":{"view_html":"https://pith.science/pith/JZFBZH74GSE6SSYIU6LZULQUSC","download_json":"https://pith.science/pith/JZFBZH74GSE6SSYIU6LZULQUSC.json","view_paper":"https://pith.science/paper/JZFBZH74","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.00645&json=true","fetch_graph":"https://pith.science/api/pith-number/JZFBZH74GSE6SSYIU6LZULQUSC/graph.json","fetch_events":"https://pith.science/api/pith-number/JZFBZH74GSE6SSYIU6LZULQUSC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JZFBZH74GSE6SSYIU6LZULQUSC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JZFBZH74GSE6SSYIU6LZULQUSC/action/storage_attestation","attest_author":"https://pith.science/pith/JZFBZH74GSE6SSYIU6LZULQUSC/action/author_attestation","sign_citation":"https://pith.science/pith/JZFBZH74GSE6SSYIU6LZULQUSC/action/citation_signature","submit_replication":"https://pith.science/pith/JZFBZH74GSE6SSYIU6LZULQUSC/action/replication_record"}},"created_at":"2026-07-05T08:27:26.521479+00:00","updated_at":"2026-07-05T08:27:26.521479+00:00"}