{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:USLONIBFBKNHJZFHPPVXGZG7BR","short_pith_number":"pith:USLONIBF","schema_version":"1.0","canonical_sha256":"a496e6a0250a9a74e4a77beb7364df0c7a382205975917d4fe1ef7291e50b9f8","source":{"kind":"arxiv","id":"2606.01151","version":1},"attestation_state":"computed","paper":{"title":"Lagrangian Perturbation Diffusion Steering: Latent Reinforcement Learning for Generative Policies","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hikmet Simsir, Ozgur S. Oguz","submitted_at":"2026-05-31T10:40:28Z","abstract_excerpt":"Behavior cloning with high-capacity generative policies achieves strong imitation performance, but is often limited by demonstration coverage and distribution shift. Direct reinforcement learning fine-tuning can improve performance, but updating large action decoders is frequently unstable and sample inefficient. We propose Lagrangian Perturbation Diffusion Steering (LP-DS), a lightweight adaptation method that improves a frozen generative policy by learning a compact noise-space perturbation before decoding. LP-DS optimizes this perturbation with a Lagrangian trust-region objective, improving"},"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":"2606.01151","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-31T10:40:28Z","cross_cats_sorted":[],"title_canon_sha256":"9f260fa9c28358becc92f6ee1ef36bd4ec09d93e8e7d6f0185963469cf47bb42","abstract_canon_sha256":"b1104f07489ac1d85e257638c4c9019be45fbfebae7680b65baedb06c1025d5d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T02:04:25.153301Z","signature_b64":"CW39NssjqJQxnRmok0gbX5tij8ohik6zk5V2Gr0AcM+gqsLIU6LpTsBLsgz+L/greYsy5TGMWgqBWokRAg4FDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a496e6a0250a9a74e4a77beb7364df0c7a382205975917d4fe1ef7291e50b9f8","last_reissued_at":"2026-06-02T02:04:25.152885Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T02:04:25.152885Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Lagrangian Perturbation Diffusion Steering: Latent Reinforcement Learning for Generative Policies","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hikmet Simsir, Ozgur S. Oguz","submitted_at":"2026-05-31T10:40:28Z","abstract_excerpt":"Behavior cloning with high-capacity generative policies achieves strong imitation performance, but is often limited by demonstration coverage and distribution shift. Direct reinforcement learning fine-tuning can improve performance, but updating large action decoders is frequently unstable and sample inefficient. We propose Lagrangian Perturbation Diffusion Steering (LP-DS), a lightweight adaptation method that improves a frozen generative policy by learning a compact noise-space perturbation before decoding. LP-DS optimizes this perturbation with a Lagrangian trust-region objective, improving"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.01151","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/2606.01151/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":"2606.01151","created_at":"2026-06-02T02:04:25.152946+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.01151v1","created_at":"2026-06-02T02:04:25.152946+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.01151","created_at":"2026-06-02T02:04:25.152946+00:00"},{"alias_kind":"pith_short_12","alias_value":"USLONIBFBKNH","created_at":"2026-06-02T02:04:25.152946+00:00"},{"alias_kind":"pith_short_16","alias_value":"USLONIBFBKNHJZFH","created_at":"2026-06-02T02:04:25.152946+00:00"},{"alias_kind":"pith_short_8","alias_value":"USLONIBF","created_at":"2026-06-02T02:04:25.152946+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/USLONIBFBKNHJZFHPPVXGZG7BR","json":"https://pith.science/pith/USLONIBFBKNHJZFHPPVXGZG7BR.json","graph_json":"https://pith.science/api/pith-number/USLONIBFBKNHJZFHPPVXGZG7BR/graph.json","events_json":"https://pith.science/api/pith-number/USLONIBFBKNHJZFHPPVXGZG7BR/events.json","paper":"https://pith.science/paper/USLONIBF"},"agent_actions":{"view_html":"https://pith.science/pith/USLONIBFBKNHJZFHPPVXGZG7BR","download_json":"https://pith.science/pith/USLONIBFBKNHJZFHPPVXGZG7BR.json","view_paper":"https://pith.science/paper/USLONIBF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.01151&json=true","fetch_graph":"https://pith.science/api/pith-number/USLONIBFBKNHJZFHPPVXGZG7BR/graph.json","fetch_events":"https://pith.science/api/pith-number/USLONIBFBKNHJZFHPPVXGZG7BR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/USLONIBFBKNHJZFHPPVXGZG7BR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/USLONIBFBKNHJZFHPPVXGZG7BR/action/storage_attestation","attest_author":"https://pith.science/pith/USLONIBFBKNHJZFHPPVXGZG7BR/action/author_attestation","sign_citation":"https://pith.science/pith/USLONIBFBKNHJZFHPPVXGZG7BR/action/citation_signature","submit_replication":"https://pith.science/pith/USLONIBFBKNHJZFHPPVXGZG7BR/action/replication_record"}},"created_at":"2026-06-02T02:04:25.152946+00:00","updated_at":"2026-06-02T02:04:25.152946+00:00"}