{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:M6OSRUENVNBVADTE2PQRSUSC64","short_pith_number":"pith:M6OSRUEN","schema_version":"1.0","canonical_sha256":"679d28d08dab43500e64d3e1195242f71f5e2bcbb94fbea968014817cb7c5e6f","source":{"kind":"arxiv","id":"2112.09737","version":2},"attestation_state":"computed","paper":{"title":"Learning to Repair: Repairing model output errors after deployment using a dynamic memory of feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Aman Madaan, Niket Tandon, Peter Clark, Yiming Yang","submitted_at":"2021-12-16T07:01:28Z","abstract_excerpt":"Large language models (LMs), while powerful, are not immune to mistakes, but can be difficult to retrain. Our goal is for an LM to continue to improve after deployment, without retraining, using feedback from the user. Our approach pairs an LM with (i) a growing memory of cases where the user identified an output error and provided general feedback on how to correct it (ii) a corrector model, trained to translate this general feedback into specific edits to repair the model output. Given a new, unseen input, our model can then use feedback from similar, past cases to repair output errors that "},"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":"2112.09737","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-12-16T07:01:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"73c52fd5a5cdad7524868e521602b3b2056df1ac01d7bc3b52dee46453512565","abstract_canon_sha256":"c317db1c3bc8b976b3e0f188e5d35b8c4062022c833b67f8895a0a62a3f9c0e6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:21:35.893622Z","signature_b64":"ecVmR0wDtvau8r2EwrRPVi+e8gpfyE457IS3CvcuGONWqYVoczj61i5oZViG5zX/SyK8iuuizUhD9Tm2uYMNDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"679d28d08dab43500e64d3e1195242f71f5e2bcbb94fbea968014817cb7c5e6f","last_reissued_at":"2026-07-05T04:21:35.893152Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:21:35.893152Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning to Repair: Repairing model output errors after deployment using a dynamic memory of feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Aman Madaan, Niket Tandon, Peter Clark, Yiming Yang","submitted_at":"2021-12-16T07:01:28Z","abstract_excerpt":"Large language models (LMs), while powerful, are not immune to mistakes, but can be difficult to retrain. Our goal is for an LM to continue to improve after deployment, without retraining, using feedback from the user. Our approach pairs an LM with (i) a growing memory of cases where the user identified an output error and provided general feedback on how to correct it (ii) a corrector model, trained to translate this general feedback into specific edits to repair the model output. Given a new, unseen input, our model can then use feedback from similar, past cases to repair output errors that "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2112.09737","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/2112.09737/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":"2112.09737","created_at":"2026-07-05T04:21:35.893206+00:00"},{"alias_kind":"arxiv_version","alias_value":"2112.09737v2","created_at":"2026-07-05T04:21:35.893206+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2112.09737","created_at":"2026-07-05T04:21:35.893206+00:00"},{"alias_kind":"pith_short_12","alias_value":"M6OSRUENVNBV","created_at":"2026-07-05T04:21:35.893206+00:00"},{"alias_kind":"pith_short_16","alias_value":"M6OSRUENVNBVADTE","created_at":"2026-07-05T04:21:35.893206+00:00"},{"alias_kind":"pith_short_8","alias_value":"M6OSRUEN","created_at":"2026-07-05T04:21:35.893206+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2403.07974","citing_title":"LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code","ref_index":136,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/M6OSRUENVNBVADTE2PQRSUSC64","json":"https://pith.science/pith/M6OSRUENVNBVADTE2PQRSUSC64.json","graph_json":"https://pith.science/api/pith-number/M6OSRUENVNBVADTE2PQRSUSC64/graph.json","events_json":"https://pith.science/api/pith-number/M6OSRUENVNBVADTE2PQRSUSC64/events.json","paper":"https://pith.science/paper/M6OSRUEN"},"agent_actions":{"view_html":"https://pith.science/pith/M6OSRUENVNBVADTE2PQRSUSC64","download_json":"https://pith.science/pith/M6OSRUENVNBVADTE2PQRSUSC64.json","view_paper":"https://pith.science/paper/M6OSRUEN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2112.09737&json=true","fetch_graph":"https://pith.science/api/pith-number/M6OSRUENVNBVADTE2PQRSUSC64/graph.json","fetch_events":"https://pith.science/api/pith-number/M6OSRUENVNBVADTE2PQRSUSC64/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/M6OSRUENVNBVADTE2PQRSUSC64/action/timestamp_anchor","attest_storage":"https://pith.science/pith/M6OSRUENVNBVADTE2PQRSUSC64/action/storage_attestation","attest_author":"https://pith.science/pith/M6OSRUENVNBVADTE2PQRSUSC64/action/author_attestation","sign_citation":"https://pith.science/pith/M6OSRUENVNBVADTE2PQRSUSC64/action/citation_signature","submit_replication":"https://pith.science/pith/M6OSRUENVNBVADTE2PQRSUSC64/action/replication_record"}},"created_at":"2026-07-05T04:21:35.893206+00:00","updated_at":"2026-07-05T04:21:35.893206+00:00"}