{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:P3BZ54GPSKAE6SHSXHUAAMB6WD","short_pith_number":"pith:P3BZ54GP","schema_version":"1.0","canonical_sha256":"7ec39ef0cf92804f48f2b9e800303eb0c91796e15df011fdd6f4d0981acd9a5c","source":{"kind":"arxiv","id":"2608.03872","version":1},"attestation_state":"computed","paper":{"title":"EvoHIL: Self-Evolving Reward and Flow-Matched Policy Optimization for Robust Human-in-the-Loop Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Bin Zheng, Jiahao Tu, Jia Pan, Jinzhuo Peng, Kai Liu, Ke Wang, Shuoqin Zhang, Tongtong Cheng, Xiru Gao, Zhe Hu","submitted_at":"2026-08-04T16:12:05Z","abstract_excerpt":"Human-in-the-loop reinforcement learning (HIL-RL) enables robots to learn contact-rich manipulation from limited real-world interaction, but deployment exposes three coupled limitations: static visual reward models fail under scene changes; independently sampled actions cause temporally inconsistent motion; and vision-based policies remain sensitive to appearance shifts. We present EvoHIL, a unified framework that adapts the reward model, action generator, and visual do main within a staged human-in-the-loop learning process. First, self-evolving reward (SER) adapts the success classifier from"},"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":"2608.03872","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-08-04T16:12:05Z","cross_cats_sorted":[],"title_canon_sha256":"3ec964744fb72b90d22e2a6d4a7ed18ea9e8feff357fd073b2ba572086494a08","abstract_canon_sha256":"37546a0448cab3f77265b62115c1eed856c1403d224b7770737486ca3b7c89d6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-05T01:37:17.735735Z","signature_b64":"PGp6IIlzP1SGQAEsV5FFO7Yef2huBgnjgLQnNPwWLE/NI7I2u/xekyuzAtu/BNTTFV8B/z1knrEipzu+1gVFCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7ec39ef0cf92804f48f2b9e800303eb0c91796e15df011fdd6f4d0981acd9a5c","last_reissued_at":"2026-08-05T01:37:17.734107Z","signature_status":"signed_v1","first_computed_at":"2026-08-05T01:37:17.734107Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"EvoHIL: Self-Evolving Reward and Flow-Matched Policy Optimization for Robust Human-in-the-Loop Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Bin Zheng, Jiahao Tu, Jia Pan, Jinzhuo Peng, Kai Liu, Ke Wang, Shuoqin Zhang, Tongtong Cheng, Xiru Gao, Zhe Hu","submitted_at":"2026-08-04T16:12:05Z","abstract_excerpt":"Human-in-the-loop reinforcement learning (HIL-RL) enables robots to learn contact-rich manipulation from limited real-world interaction, but deployment exposes three coupled limitations: static visual reward models fail under scene changes; independently sampled actions cause temporally inconsistent motion; and vision-based policies remain sensitive to appearance shifts. We present EvoHIL, a unified framework that adapts the reward model, action generator, and visual do main within a staged human-in-the-loop learning process. First, self-evolving reward (SER) adapts the success classifier from"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.03872","kind":"arxiv","version":1},"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/2608.03872/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":"2608.03872","created_at":"2026-08-05T01:37:17.734723+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.03872v1","created_at":"2026-08-05T01:37:17.734723+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.03872","created_at":"2026-08-05T01:37:17.734723+00:00"},{"alias_kind":"pith_short_12","alias_value":"P3BZ54GPSKAE","created_at":"2026-08-05T01:37:17.734723+00:00"},{"alias_kind":"pith_short_16","alias_value":"P3BZ54GPSKAE6SHS","created_at":"2026-08-05T01:37:17.734723+00:00"},{"alias_kind":"pith_short_8","alias_value":"P3BZ54GP","created_at":"2026-08-05T01:37:17.734723+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/P3BZ54GPSKAE6SHSXHUAAMB6WD","json":"https://pith.science/pith/P3BZ54GPSKAE6SHSXHUAAMB6WD.json","graph_json":"https://pith.science/api/pith-number/P3BZ54GPSKAE6SHSXHUAAMB6WD/graph.json","events_json":"https://pith.science/api/pith-number/P3BZ54GPSKAE6SHSXHUAAMB6WD/events.json","paper":"https://pith.science/paper/P3BZ54GP"},"agent_actions":{"view_html":"https://pith.science/pith/P3BZ54GPSKAE6SHSXHUAAMB6WD","download_json":"https://pith.science/pith/P3BZ54GPSKAE6SHSXHUAAMB6WD.json","view_paper":"https://pith.science/paper/P3BZ54GP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.03872&json=true","fetch_graph":"https://pith.science/api/pith-number/P3BZ54GPSKAE6SHSXHUAAMB6WD/graph.json","fetch_events":"https://pith.science/api/pith-number/P3BZ54GPSKAE6SHSXHUAAMB6WD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/P3BZ54GPSKAE6SHSXHUAAMB6WD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/P3BZ54GPSKAE6SHSXHUAAMB6WD/action/storage_attestation","attest_author":"https://pith.science/pith/P3BZ54GPSKAE6SHSXHUAAMB6WD/action/author_attestation","sign_citation":"https://pith.science/pith/P3BZ54GPSKAE6SHSXHUAAMB6WD/action/citation_signature","submit_replication":"https://pith.science/pith/P3BZ54GPSKAE6SHSXHUAAMB6WD/action/replication_record"}},"created_at":"2026-08-05T01:37:17.734723+00:00","updated_at":"2026-08-05T01:37:17.734723+00:00"}