{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:NPLNJJLSEF7SXWTUYD54X5ZXFF","short_pith_number":"pith:NPLNJJLS","schema_version":"1.0","canonical_sha256":"6bd6d4a572217f2bda74c0fbcbf737295077a0a2fa751dfd67be81189995a040","source":{"kind":"arxiv","id":"2210.14215","version":1},"attestation_state":"computed","paper":{"title":"In-context Reinforcement Learning with Algorithm Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Angelos Filos, DJ Strouse, Emilio Parisotto, Ethan Brooks, Himanshu Sahni, Junhyuk Oh, Luyu Wang, Maxime Gazeau, Michael Laskin, Richie Steigerwald, Satinder Singh, Stephen Spencer, Steven Hansen, Volodymyr Mnih","submitted_at":"2022-10-25T17:57:49Z","abstract_excerpt":"We propose Algorithm Distillation (AD), a method for distilling reinforcement learning (RL) algorithms into neural networks by modeling their training histories with a causal sequence model. Algorithm Distillation treats learning to reinforcement learn as an across-episode sequential prediction problem. A dataset of learning histories is generated by a source RL algorithm, and then a causal transformer is trained by autoregressively predicting actions given their preceding learning histories as context. Unlike sequential policy prediction architectures that distill post-learning or expert sequ"},"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":"2210.14215","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-10-25T17:57:49Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e8511c400467c7c35dcc8495c9d073da266e30f69e495dc855a0c82b9a5a4ee4","abstract_canon_sha256":"0f7f8eb5360a136bb17ae0151205088a38bf1dabc63919bc7153c5aeb76b24fb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:10:27.534646Z","signature_b64":"Bml2Pjf/6zyUf+S08EYvqucSHyTYeHjNfm7ioliKJkw5m3FIkKtWi014Nec5NbUcFILXjgdzTFJkCftzorzOBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6bd6d4a572217f2bda74c0fbcbf737295077a0a2fa751dfd67be81189995a040","last_reissued_at":"2026-07-05T05:10:27.534150Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:10:27.534150Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"In-context Reinforcement Learning with Algorithm Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Angelos Filos, DJ Strouse, Emilio Parisotto, Ethan Brooks, Himanshu Sahni, Junhyuk Oh, Luyu Wang, Maxime Gazeau, Michael Laskin, Richie Steigerwald, Satinder Singh, Stephen Spencer, Steven Hansen, Volodymyr Mnih","submitted_at":"2022-10-25T17:57:49Z","abstract_excerpt":"We propose Algorithm Distillation (AD), a method for distilling reinforcement learning (RL) algorithms into neural networks by modeling their training histories with a causal sequence model. Algorithm Distillation treats learning to reinforcement learn as an across-episode sequential prediction problem. A dataset of learning histories is generated by a source RL algorithm, and then a causal transformer is trained by autoregressively predicting actions given their preceding learning histories as context. Unlike sequential policy prediction architectures that distill post-learning or expert sequ"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2210.14215","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/2210.14215/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":"2210.14215","created_at":"2026-07-05T05:10:27.534211+00:00"},{"alias_kind":"arxiv_version","alias_value":"2210.14215v1","created_at":"2026-07-05T05:10:27.534211+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2210.14215","created_at":"2026-07-05T05:10:27.534211+00:00"},{"alias_kind":"pith_short_12","alias_value":"NPLNJJLSEF7S","created_at":"2026-07-05T05:10:27.534211+00:00"},{"alias_kind":"pith_short_16","alias_value":"NPLNJJLSEF7SXWTU","created_at":"2026-07-05T05:10:27.534211+00:00"},{"alias_kind":"pith_short_8","alias_value":"NPLNJJLS","created_at":"2026-07-05T05:10:27.534211+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":15,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22719","citing_title":"Leakage-Aware Benchmarking of LLM Forecasting: Real-Time Nowcasts as the Decision-Time Input for Macro Factor Ranking","ref_index":92,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18812","citing_title":"Reinforcement Learning Foundation Models Should Already Be A Thing","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.17477","citing_title":"Theoretical Grounding of Out-Of-Distribution Detection With Reinforcement Learning Optimizer","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18315","citing_title":"Ghost Attractor Networks: Basin-Structured Dynamical Decoders for Closed-Loop Sequential Generation","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07568","citing_title":"A Systematic Study of Behavioral Cloning for Scientific Data Annotation","ref_index":106,"is_internal_anchor":false},{"citing_arxiv_id":"2211.15661","citing_title":"What learning algorithm is in-context learning? Investigations with linear models","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2309.16797","citing_title":"Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolution","ref_index":192,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12288","citing_title":"TokenRatio: Principled Token-Level Preference Optimization via Ratio Matching","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2507.21046","citing_title":"A Survey of Self-Evolving Agents: What, When, How, and Where to Evolve on the Path to Artificial Super Intelligence","ref_index":131,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12288","citing_title":"TokenRatio: Principled Token-Level Preference Optimization via Ratio Matching","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09150","citing_title":"AlphaExploitem: Going Beyond the Nash Equilibrium in Poker by Learning to Exploit Suboptimal Play","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09009","citing_title":"Large Language Models for Sequential Decision-Making: Improving In-Context Learning via Supervised Fine-Tuning","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06992","citing_title":"Why Does Agentic Safety Fail to Generalize Across Tasks?","ref_index":62,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05859","citing_title":"When Do We Need LLMs? A Diagnostic for Language-Driven Bandits","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04291","citing_title":"Leveraging Pretrained Language Models as Energy Functions for Glauber Dynamics Text Diffusion","ref_index":60,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NPLNJJLSEF7SXWTUYD54X5ZXFF","json":"https://pith.science/pith/NPLNJJLSEF7SXWTUYD54X5ZXFF.json","graph_json":"https://pith.science/api/pith-number/NPLNJJLSEF7SXWTUYD54X5ZXFF/graph.json","events_json":"https://pith.science/api/pith-number/NPLNJJLSEF7SXWTUYD54X5ZXFF/events.json","paper":"https://pith.science/paper/NPLNJJLS"},"agent_actions":{"view_html":"https://pith.science/pith/NPLNJJLSEF7SXWTUYD54X5ZXFF","download_json":"https://pith.science/pith/NPLNJJLSEF7SXWTUYD54X5ZXFF.json","view_paper":"https://pith.science/paper/NPLNJJLS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2210.14215&json=true","fetch_graph":"https://pith.science/api/pith-number/NPLNJJLSEF7SXWTUYD54X5ZXFF/graph.json","fetch_events":"https://pith.science/api/pith-number/NPLNJJLSEF7SXWTUYD54X5ZXFF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NPLNJJLSEF7SXWTUYD54X5ZXFF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NPLNJJLSEF7SXWTUYD54X5ZXFF/action/storage_attestation","attest_author":"https://pith.science/pith/NPLNJJLSEF7SXWTUYD54X5ZXFF/action/author_attestation","sign_citation":"https://pith.science/pith/NPLNJJLSEF7SXWTUYD54X5ZXFF/action/citation_signature","submit_replication":"https://pith.science/pith/NPLNJJLSEF7SXWTUYD54X5ZXFF/action/replication_record"}},"created_at":"2026-07-05T05:10:27.534211+00:00","updated_at":"2026-07-05T05:10:27.534211+00:00"}