{"paper":{"title":"CAB: Accelerating Flow and Diffusion Sampling via Rectification and Corrected Adams-Bashforth","license":"http://creativecommons.org/licenses/by/4.0/","headline":"CAB unifies flow and diffusion sampling by rectifying dynamics to one coordinate system then applying a corrected multistep predictor that needs no extra model calls.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Anuska Roy, Pravin Nair","submitted_at":"2026-05-16T01:16:18Z","abstract_excerpt":"Flow and diffusion models achieve high-fidelity, high-resolution image synthesis, but often require many function evaluations (NFEs) at sampling time. Existing acceleration methods either require additional training through distillation or rely on training-free high-order solvers, and both can degrade sample quality at low NFE budgets. We propose CAB (Corrected Adams-Bashforth), a training-free sampler that accelerates both flow and diffusion models. CAB first transforms the sampling dynamics to a common rectified coordinate system, and then applies a multistep Adams-Bashforth predictor augmen"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The resulting method is simple, has the same algorithmic form across model classes, and has at least third-order local truncation error and second-order global error. Experiments on pretrained flow and diffusion models show that CAB improves quality-NFE trade-offs in the low-step regime of 6-20 NFEs.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the rectification step transforms the sampling dynamics of both flow and diffusion models into a common coordinate system where the multistep Adams-Bashforth predictor with correction term can be applied uniformly without introducing instability or order reduction.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CAB accelerates flow and diffusion sampling via rectification to a shared coordinate system followed by a corrected Adams-Bashforth predictor that achieves third-order local truncation error while using no additional NFEs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"CAB unifies flow and diffusion sampling by rectifying dynamics to one coordinate system then applying a corrected multistep predictor that needs no extra model calls.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"6d8bd74048c47b4db90fffda380e73883ab3e4b4afd05746608a894dbe6e3dea"},"source":{"id":"2605.16736","kind":"arxiv","version":1},"verdict":{"id":"399aedcf-40e9-4489-8a91-ef9cc5f427d7","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T21:37:34.430104Z","strongest_claim":"The resulting method is simple, has the same algorithmic form across model classes, and has at least third-order local truncation error and second-order global error. Experiments on pretrained flow and diffusion models show that CAB improves quality-NFE trade-offs in the low-step regime of 6-20 NFEs.","one_line_summary":"CAB accelerates flow and diffusion sampling via rectification to a shared coordinate system followed by a corrected Adams-Bashforth predictor that achieves third-order local truncation error while using no additional NFEs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the rectification step transforms the sampling dynamics of both flow and diffusion models into a common coordinate system where the multistep Adams-Bashforth predictor with correction term can be applied uniformly without introducing instability or order reduction.","pith_extraction_headline":"CAB unifies flow and diffusion sampling by rectifying dynamics to one coordinate system then applying a corrected multistep predictor that needs no extra model calls."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.16736/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T22:01:19.904601Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T21:51:05.203058Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T19:01:56.338124Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.466479Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"632004d6041db3e6ee0b96ac3ba02e2b26aa2d0fc836d84a53033d0a2fbd4a96"},"references":{"count":82,"sample":[{"doi":"","year":null,"title":"Denoising diffusion probabilistic models , author=. Proc. Advances in neural information processing systems , volume=","work_id":"cbe3c85b-ddc7-450c-88f8-57357a7c4155","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Flow Matching for Generative Modeling , author=. Proc. International Conference on Learning Representations , year=","work_id":"63bd3ecc-3129-45b5-a88e-c5cc5b739a8c","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Video Diffusion Models , author =. Proc. Advances in neural information processing systems , volume=","work_id":"7ec8f3f0-529a-463b-9053-e25f568c7d76","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=","work_id":"5adfb806-8536-4b3f-8b12-95855a5e540d","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"NeurIPS 2021 Workshop on Deep Generative Models and Downstream Applications , year =","work_id":"b67969d6-7ae4-4413-8926-c33d2327a0b2","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":82,"snapshot_sha256":"e27428deec63559f1c981e2ba18a81869832d2372413302dfaeb7712b0257790","internal_anchors":2},"formal_canon":{"evidence_count":2,"snapshot_sha256":"24031508b6e909224025959dd11d9307c5c91310884a106f30647897613f40d7"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}