{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:B357M62R5Q7MF7Z25QIU3FVKN6","short_pith_number":"pith:B357M62R","schema_version":"1.0","canonical_sha256":"0efbf67b51ec3ec2ff3aec114d96aa6fbe7be788701cc5084f3ff7d7c75e7939","source":{"kind":"arxiv","id":"2503.13817","version":1},"attestation_state":"computed","paper":{"title":"VARP: Reinforcement Learning from Vision-Language Model Feedback with Agent Regularized Preferences","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.HC","cs.LG","cs.RO"],"primary_cat":"cs.AI","authors_text":"Amisha Bhaskar, Amrit Bedi, Anukriti Singh, Peihong Yu, Pratap Tokekar, Ruthwik Dasyam, Souradip Chakraborty","submitted_at":"2025-03-18T01:51:27Z","abstract_excerpt":"Designing reward functions for continuous-control robotics often leads to subtle misalignments or reward hacking, especially in complex tasks. Preference-based RL mitigates some of these pitfalls by learning rewards from comparative feedback rather than hand-crafted signals, yet scaling human annotations remains challenging. Recent work uses Vision-Language Models (VLMs) to automate preference labeling, but a single final-state image generally fails to capture the agent's full motion. In this paper, we present a two-part solution that both improves feedback accuracy and better aligns reward le"},"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":"2503.13817","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-03-18T01:51:27Z","cross_cats_sorted":["cs.HC","cs.LG","cs.RO"],"title_canon_sha256":"6ce926aeab3ced030b4359e6310e8cafec99ed034ea3daf1cdc7f658e73b3095","abstract_canon_sha256":"78c4cb09db3cbaacb5322c1e5e90d1c6b4aa4dacb21c3db576951e714fa89598"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:33:19.112447Z","signature_b64":"gLoCurt8MI7qItqRWpANGtTBpAP+qZN1M9DClLHFT/L+ToNa1zFEKd89W+KxRdmvIOzb57JUCx05M55azylbDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0efbf67b51ec3ec2ff3aec114d96aa6fbe7be788701cc5084f3ff7d7c75e7939","last_reissued_at":"2026-07-05T10:33:19.111953Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:33:19.111953Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"VARP: Reinforcement Learning from Vision-Language Model Feedback with Agent Regularized Preferences","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.HC","cs.LG","cs.RO"],"primary_cat":"cs.AI","authors_text":"Amisha Bhaskar, Amrit Bedi, Anukriti Singh, Peihong Yu, Pratap Tokekar, Ruthwik Dasyam, Souradip Chakraborty","submitted_at":"2025-03-18T01:51:27Z","abstract_excerpt":"Designing reward functions for continuous-control robotics often leads to subtle misalignments or reward hacking, especially in complex tasks. Preference-based RL mitigates some of these pitfalls by learning rewards from comparative feedback rather than hand-crafted signals, yet scaling human annotations remains challenging. Recent work uses Vision-Language Models (VLMs) to automate preference labeling, but a single final-state image generally fails to capture the agent's full motion. In this paper, we present a two-part solution that both improves feedback accuracy and better aligns reward le"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.13817","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/2503.13817/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":"2503.13817","created_at":"2026-07-05T10:33:19.112011+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.13817v1","created_at":"2026-07-05T10:33:19.112011+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.13817","created_at":"2026-07-05T10:33:19.112011+00:00"},{"alias_kind":"pith_short_12","alias_value":"B357M62R5Q7M","created_at":"2026-07-05T10:33:19.112011+00:00"},{"alias_kind":"pith_short_16","alias_value":"B357M62R5Q7MF7Z2","created_at":"2026-07-05T10:33:19.112011+00:00"},{"alias_kind":"pith_short_8","alias_value":"B357M62R","created_at":"2026-07-05T10:33:19.112011+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.12748","citing_title":"Scaling Representation Diversity: Modulated Attention and Reconstructive Regularization for Visual Grounding","ref_index":204,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/B357M62R5Q7MF7Z25QIU3FVKN6","json":"https://pith.science/pith/B357M62R5Q7MF7Z25QIU3FVKN6.json","graph_json":"https://pith.science/api/pith-number/B357M62R5Q7MF7Z25QIU3FVKN6/graph.json","events_json":"https://pith.science/api/pith-number/B357M62R5Q7MF7Z25QIU3FVKN6/events.json","paper":"https://pith.science/paper/B357M62R"},"agent_actions":{"view_html":"https://pith.science/pith/B357M62R5Q7MF7Z25QIU3FVKN6","download_json":"https://pith.science/pith/B357M62R5Q7MF7Z25QIU3FVKN6.json","view_paper":"https://pith.science/paper/B357M62R","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.13817&json=true","fetch_graph":"https://pith.science/api/pith-number/B357M62R5Q7MF7Z25QIU3FVKN6/graph.json","fetch_events":"https://pith.science/api/pith-number/B357M62R5Q7MF7Z25QIU3FVKN6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B357M62R5Q7MF7Z25QIU3FVKN6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B357M62R5Q7MF7Z25QIU3FVKN6/action/storage_attestation","attest_author":"https://pith.science/pith/B357M62R5Q7MF7Z25QIU3FVKN6/action/author_attestation","sign_citation":"https://pith.science/pith/B357M62R5Q7MF7Z25QIU3FVKN6/action/citation_signature","submit_replication":"https://pith.science/pith/B357M62R5Q7MF7Z25QIU3FVKN6/action/replication_record"}},"created_at":"2026-07-05T10:33:19.112011+00:00","updated_at":"2026-07-05T10:33:19.112011+00:00"}