{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:B4PBXILFMETNCJTLJQY5ATBTIE","short_pith_number":"pith:B4PBXILF","schema_version":"1.0","canonical_sha256":"0f1e1ba1656126d1266b4c31d04c33412f04d01f74b5bd28e47588ebed03ea50","source":{"kind":"arxiv","id":"2211.04878","version":1},"attestation_state":"computed","paper":{"title":"Foundation Models for Semantic Novelty in Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Danfei Xu, Marco Pavone, Peter Karkus, Tarun Gupta, Tong Che","submitted_at":"2022-11-09T13:34:45Z","abstract_excerpt":"Effectively exploring the environment is a key challenge in reinforcement learning (RL). We address this challenge by defining a novel intrinsic reward based on a foundation model, such as contrastive language image pretraining (CLIP), which can encode a wealth of domain-independent semantic visual-language knowledge about the world. Specifically, our intrinsic reward is defined based on pre-trained CLIP embeddings without any fine-tuning or learning on the target RL task. We demonstrate that CLIP-based intrinsic rewards can drive exploration towards semantically meaningful states and outperfo"},"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":"2211.04878","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-11-09T13:34:45Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"736cdb18e639a5dcb850225eddecd1bf21d691cc3548ebfdd59f777c0b5cac84","abstract_canon_sha256":"3909ff753807c78f92ce0397a7c3f9004159bc4ed950c580f0cf2cae08d3d6cc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:14:42.852103Z","signature_b64":"xSSrOB4f/6FGd9LMNE6MnmBulEoeTh5ZuOR0q5ksilqM8tJ1pfP3wrTfm6PwOzsNAkrJZTTuz4zewLdeDSs5BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0f1e1ba1656126d1266b4c31d04c33412f04d01f74b5bd28e47588ebed03ea50","last_reissued_at":"2026-07-05T05:14:42.851722Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:14:42.851722Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Foundation Models for Semantic Novelty in Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Danfei Xu, Marco Pavone, Peter Karkus, Tarun Gupta, Tong Che","submitted_at":"2022-11-09T13:34:45Z","abstract_excerpt":"Effectively exploring the environment is a key challenge in reinforcement learning (RL). We address this challenge by defining a novel intrinsic reward based on a foundation model, such as contrastive language image pretraining (CLIP), which can encode a wealth of domain-independent semantic visual-language knowledge about the world. Specifically, our intrinsic reward is defined based on pre-trained CLIP embeddings without any fine-tuning or learning on the target RL task. We demonstrate that CLIP-based intrinsic rewards can drive exploration towards semantically meaningful states and outperfo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2211.04878","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/2211.04878/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":"2211.04878","created_at":"2026-07-05T05:14:42.851772+00:00"},{"alias_kind":"arxiv_version","alias_value":"2211.04878v1","created_at":"2026-07-05T05:14:42.851772+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2211.04878","created_at":"2026-07-05T05:14:42.851772+00:00"},{"alias_kind":"pith_short_12","alias_value":"B4PBXILFMETN","created_at":"2026-07-05T05:14:42.851772+00:00"},{"alias_kind":"pith_short_16","alias_value":"B4PBXILFMETNCJTL","created_at":"2026-07-05T05:14:42.851772+00:00"},{"alias_kind":"pith_short_8","alias_value":"B4PBXILF","created_at":"2026-07-05T05:14:42.851772+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.03234","citing_title":"Right Makes Might: Aligning Verified Hidden States Empowers RL Reasoning","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2309.16797","citing_title":"Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolution","ref_index":126,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/B4PBXILFMETNCJTLJQY5ATBTIE","json":"https://pith.science/pith/B4PBXILFMETNCJTLJQY5ATBTIE.json","graph_json":"https://pith.science/api/pith-number/B4PBXILFMETNCJTLJQY5ATBTIE/graph.json","events_json":"https://pith.science/api/pith-number/B4PBXILFMETNCJTLJQY5ATBTIE/events.json","paper":"https://pith.science/paper/B4PBXILF"},"agent_actions":{"view_html":"https://pith.science/pith/B4PBXILFMETNCJTLJQY5ATBTIE","download_json":"https://pith.science/pith/B4PBXILFMETNCJTLJQY5ATBTIE.json","view_paper":"https://pith.science/paper/B4PBXILF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2211.04878&json=true","fetch_graph":"https://pith.science/api/pith-number/B4PBXILFMETNCJTLJQY5ATBTIE/graph.json","fetch_events":"https://pith.science/api/pith-number/B4PBXILFMETNCJTLJQY5ATBTIE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B4PBXILFMETNCJTLJQY5ATBTIE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B4PBXILFMETNCJTLJQY5ATBTIE/action/storage_attestation","attest_author":"https://pith.science/pith/B4PBXILFMETNCJTLJQY5ATBTIE/action/author_attestation","sign_citation":"https://pith.science/pith/B4PBXILFMETNCJTLJQY5ATBTIE/action/citation_signature","submit_replication":"https://pith.science/pith/B4PBXILFMETNCJTLJQY5ATBTIE/action/replication_record"}},"created_at":"2026-07-05T05:14:42.851772+00:00","updated_at":"2026-07-05T05:14:42.851772+00:00"}