{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:23TAMHIQ3APVWUOJGCN5OUO7VW","short_pith_number":"pith:23TAMHIQ","schema_version":"1.0","canonical_sha256":"d6e6061d10d81f5b51c9309bd751dfadbfdc99ed748b9976d1779f8a45c7c435","source":{"kind":"arxiv","id":"2503.09722","version":4},"attestation_state":"computed","paper":{"title":"The Pitfalls of Imitation Learning when Actions are Continuous","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.SY","eess.SY","stat.ML"],"primary_cat":"cs.LG","authors_text":"Ali Jadbabaie, Daniel Pfrommer, Max Simchowitz","submitted_at":"2025-03-12T18:11:37Z","abstract_excerpt":"We study the problem of imitating an expert demonstrator in a discrete-time, continuous state-and-action control system. We show that, even if the dynamics satisfy a control-theoretic property called exponential stability (i.e. the effects of perturbations decay exponentially quickly), and the expert is smooth and deterministic, any smooth, deterministic imitator policy necessarily suffers error on execution that is exponentially larger, as a function of problem horizon, than the error under the distribution of expert training data. Our negative result applies to any algorithm which learns sol"},"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":"2503.09722","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-03-12T18:11:37Z","cross_cats_sorted":["cs.SY","eess.SY","stat.ML"],"title_canon_sha256":"4bed1f80b5a48817cbce235f257d77a1b5f129aa640be75e79b37f092be501d2","abstract_canon_sha256":"3b501b7b960c401a99a8d11d0a66642ce0bacb9348f56437e8871171ce01288e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:43:31.115693Z","signature_b64":"fw+/AarpfYR6bbHJYtAbAH7mqBzqbwzeTHX1d6Hsk6gjmAvZ3oyMvOCH5iwreSwhldWDVvIPuD2s0MB0pr96BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d6e6061d10d81f5b51c9309bd751dfadbfdc99ed748b9976d1779f8a45c7c435","last_reissued_at":"2026-07-05T11:43:31.115173Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:43:31.115173Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Pitfalls of Imitation Learning when Actions are Continuous","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.SY","eess.SY","stat.ML"],"primary_cat":"cs.LG","authors_text":"Ali Jadbabaie, Daniel Pfrommer, Max Simchowitz","submitted_at":"2025-03-12T18:11:37Z","abstract_excerpt":"We study the problem of imitating an expert demonstrator in a discrete-time, continuous state-and-action control system. We show that, even if the dynamics satisfy a control-theoretic property called exponential stability (i.e. the effects of perturbations decay exponentially quickly), and the expert is smooth and deterministic, any smooth, deterministic imitator policy necessarily suffers error on execution that is exponentially larger, as a function of problem horizon, than the error under the distribution of expert training data. Our negative result applies to any algorithm which learns sol"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.09722","kind":"arxiv","version":4},"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/2503.09722/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":"2503.09722","created_at":"2026-07-05T11:43:31.115232+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.09722v4","created_at":"2026-07-05T11:43:31.115232+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.09722","created_at":"2026-07-05T11:43:31.115232+00:00"},{"alias_kind":"pith_short_12","alias_value":"23TAMHIQ3APV","created_at":"2026-07-05T11:43:31.115232+00:00"},{"alias_kind":"pith_short_16","alias_value":"23TAMHIQ3APVWUOJ","created_at":"2026-07-05T11:43:31.115232+00:00"},{"alias_kind":"pith_short_8","alias_value":"23TAMHIQ","created_at":"2026-07-05T11:43:31.115232+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.22493","citing_title":"Understanding Multimodal Failure in Action-Chunking Behavioral Cloning","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10165","citing_title":"MoRI: Mixture of RL and IL Experts for Long-Horizon Manipulation Tasks","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/23TAMHIQ3APVWUOJGCN5OUO7VW","json":"https://pith.science/pith/23TAMHIQ3APVWUOJGCN5OUO7VW.json","graph_json":"https://pith.science/api/pith-number/23TAMHIQ3APVWUOJGCN5OUO7VW/graph.json","events_json":"https://pith.science/api/pith-number/23TAMHIQ3APVWUOJGCN5OUO7VW/events.json","paper":"https://pith.science/paper/23TAMHIQ"},"agent_actions":{"view_html":"https://pith.science/pith/23TAMHIQ3APVWUOJGCN5OUO7VW","download_json":"https://pith.science/pith/23TAMHIQ3APVWUOJGCN5OUO7VW.json","view_paper":"https://pith.science/paper/23TAMHIQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.09722&json=true","fetch_graph":"https://pith.science/api/pith-number/23TAMHIQ3APVWUOJGCN5OUO7VW/graph.json","fetch_events":"https://pith.science/api/pith-number/23TAMHIQ3APVWUOJGCN5OUO7VW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/23TAMHIQ3APVWUOJGCN5OUO7VW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/23TAMHIQ3APVWUOJGCN5OUO7VW/action/storage_attestation","attest_author":"https://pith.science/pith/23TAMHIQ3APVWUOJGCN5OUO7VW/action/author_attestation","sign_citation":"https://pith.science/pith/23TAMHIQ3APVWUOJGCN5OUO7VW/action/citation_signature","submit_replication":"https://pith.science/pith/23TAMHIQ3APVWUOJGCN5OUO7VW/action/replication_record"}},"created_at":"2026-07-05T11:43:31.115232+00:00","updated_at":"2026-07-05T11:43:31.115232+00:00"}