{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FIT3V26QTXAUY2OE6CLKPFSPXH","short_pith_number":"pith:FIT3V26Q","schema_version":"1.0","canonical_sha256":"2a27baebd09dc14c69c4f096a7964fb9e613559e5dea8dc674c7af8e381b89d1","source":{"kind":"arxiv","id":"2501.06848","version":5},"attestation_state":"computed","paper":{"title":"A General Framework for Inference-time Scaling and Steering of Diffusion Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Kathleen McKeown, Mengye Ren, Raghav Singhal, Rajesh Ranganath, Ryan Teehan, Zachary Horvitz, Zhou Yu","submitted_at":"2025-01-12T15:34:24Z","abstract_excerpt":"Diffusion models produce impressive results in modalities ranging from images and video to protein design and text. However, generating samples with user-specified properties remains a challenge. Recent research proposes fine-tuning models to maximize rewards that capture desired properties, but these methods require expensive training and are prone to mode collapse. In this work, we present Feynman-Kac (FK) steering, an inference-time framework for steering diffusion models with reward functions. FK steering works by sampling a system of multiple interacting diffusion processes, called partic"},"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":"2501.06848","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-01-12T15:34:24Z","cross_cats_sorted":["cs.CL","cs.CV"],"title_canon_sha256":"7b07b4ce9ad4abe83a88076c7e68885c1fab993210f8ec7d77546056eec4af0b","abstract_canon_sha256":"9f1a8fd8160d6356fdbae54e440f05b6ea4175405b82fd0122c2f55e38135030"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:39:06.513488Z","signature_b64":"WELkR1PR1OjfUUNMUt7497S5T7ObEYJ5vYZ3HL+ZqsR0zDJsscyPbZxRu5P7o59uDWU1vk2qy4dXzSEqx2SwAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2a27baebd09dc14c69c4f096a7964fb9e613559e5dea8dc674c7af8e381b89d1","last_reissued_at":"2026-07-05T11:39:06.512909Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:39:06.512909Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A General Framework for Inference-time Scaling and Steering of Diffusion Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Kathleen McKeown, Mengye Ren, Raghav Singhal, Rajesh Ranganath, Ryan Teehan, Zachary Horvitz, Zhou Yu","submitted_at":"2025-01-12T15:34:24Z","abstract_excerpt":"Diffusion models produce impressive results in modalities ranging from images and video to protein design and text. However, generating samples with user-specified properties remains a challenge. Recent research proposes fine-tuning models to maximize rewards that capture desired properties, but these methods require expensive training and are prone to mode collapse. In this work, we present Feynman-Kac (FK) steering, an inference-time framework for steering diffusion models with reward functions. FK steering works by sampling a system of multiple interacting diffusion processes, called partic"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.06848","kind":"arxiv","version":5},"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/2501.06848/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":"2501.06848","created_at":"2026-07-05T11:39:06.512976+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.06848v5","created_at":"2026-07-05T11:39:06.512976+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.06848","created_at":"2026-07-05T11:39:06.512976+00:00"},{"alias_kind":"pith_short_12","alias_value":"FIT3V26QTXAU","created_at":"2026-07-05T11:39:06.512976+00:00"},{"alias_kind":"pith_short_16","alias_value":"FIT3V26QTXAUY2OE","created_at":"2026-07-05T11:39:06.512976+00:00"},{"alias_kind":"pith_short_8","alias_value":"FIT3V26Q","created_at":"2026-07-05T11:39:06.512976+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":42,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26588","citing_title":"Inference-Time Robot Behavior Steering through Physically-Aware Reconfiguration of Task-Structure","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.21014","citing_title":"BayesFP: Posterior Estimation for Flow-Based Policies via Feynman-Kac Sampling","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2606.21395","citing_title":"Atomistic Language Models Understand and Generate Materials","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18066","citing_title":"NoiseTilt: Noise-Tilted Reverse Kernels for Diffusion Reward Alignment","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2606.13675","citing_title":"Improving Robotic Generalist Policies via Flow Reversal Steering","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12807","citing_title":"Detect, Remask, Repair: Diffusion Editing for Faithful Summarization of Evolving Contexts","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12475","citing_title":"Learning to Assist: Collaborative VLAs for Implicit Human-Robot Collaboration","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11437","citing_title":"The Power of Test-Time Training for Approximate Sampling","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08375","citing_title":"Few-step Cofolding with All-Atom Flow Maps","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00947","citing_title":"Diffeomorphic Optimization","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01144","citing_title":"Sequentially-Controlled Interactive Multi-Particle Flow-Maps for Online Feedback-Driven Search","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00647","citing_title":"Not All Prediction Targets Keep Training-Free Diffusion Guidance on the Manifold","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18745","citing_title":"SURGE: Approximation and Training Free Particle Filter for Diffusion Surrogate","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27147","citing_title":"How to Guide Your Flow: Few-Step Alignment via Flow Map Reward Guidance","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25123","citing_title":"Inference-Time Alignment of Diffusion Models via Trust-Region Iterative Twisted Sequential Monte Carlo","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18066","citing_title":"NoiseTilt: Noise-Tilted Reverse Kernels for Diffusion Reward Alignment","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30991","citing_title":"Parallel Tempering Initial Sampling in Inference-Time Reward Alignment","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23346","citing_title":"Contrastive Distribution Matching for Amortized Sequential Monte Carlo in Discrete Diffusion","ref_index":73,"is_internal_anchor":false},{"citing_arxiv_id":"2601.21484","citing_title":"ETS: Energy-Guided Test-Time Scaling for Training-Free RL Alignment","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27147","citing_title":"How to Guide Your Flow: Few-Step Alignment via Flow Map Reward Guidance","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11871","citing_title":"$h$-control: Training-Free Camera Control via Block-Conditional Gibbs Refinement","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16348","citing_title":"Flow-Direct: Feedback-Efficient and Reusable Guidance for Flow Models via Non-Parametric Guidance Field","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17244","citing_title":"Drift Flow Matching","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18745","citing_title":"SURGE: Approximation and Training Free Particle Filter for Diffusion Surrogate","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19317","citing_title":"Inference-Time Scaling in Diffusion Models through Iterative Partial Refinement","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FIT3V26QTXAUY2OE6CLKPFSPXH","json":"https://pith.science/pith/FIT3V26QTXAUY2OE6CLKPFSPXH.json","graph_json":"https://pith.science/api/pith-number/FIT3V26QTXAUY2OE6CLKPFSPXH/graph.json","events_json":"https://pith.science/api/pith-number/FIT3V26QTXAUY2OE6CLKPFSPXH/events.json","paper":"https://pith.science/paper/FIT3V26Q"},"agent_actions":{"view_html":"https://pith.science/pith/FIT3V26QTXAUY2OE6CLKPFSPXH","download_json":"https://pith.science/pith/FIT3V26QTXAUY2OE6CLKPFSPXH.json","view_paper":"https://pith.science/paper/FIT3V26Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.06848&json=true","fetch_graph":"https://pith.science/api/pith-number/FIT3V26QTXAUY2OE6CLKPFSPXH/graph.json","fetch_events":"https://pith.science/api/pith-number/FIT3V26QTXAUY2OE6CLKPFSPXH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FIT3V26QTXAUY2OE6CLKPFSPXH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FIT3V26QTXAUY2OE6CLKPFSPXH/action/storage_attestation","attest_author":"https://pith.science/pith/FIT3V26QTXAUY2OE6CLKPFSPXH/action/author_attestation","sign_citation":"https://pith.science/pith/FIT3V26QTXAUY2OE6CLKPFSPXH/action/citation_signature","submit_replication":"https://pith.science/pith/FIT3V26QTXAUY2OE6CLKPFSPXH/action/replication_record"}},"created_at":"2026-07-05T11:39:06.512976+00:00","updated_at":"2026-07-05T11:39:06.512976+00:00"}