Metraipolis › archive › dc_03fe541febe2
Risk-Calibrated Proposal Transport for Finite-Particle Diffusion Steering — arXiv:2610.04171
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.04171
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T05:26:02.867Z
- SHA-256 of the archived text
808cc160d94719b18e7f3b9c112628633aa6b832b2f09222b567f4b0dfe5b365- Stable URL of this page
https://metraipolis.com/documento/dc_03fe541febe2
How to cite
Risk-Calibrated Proposal Transport for Finite-Particle Diffusion Steering — arXiv:2610.04171 — arXiv, https://arxiv.org/abs/2610.04171 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_03fe541febe2 (Metraipolis archive dc_03fe541febe2, sha256 808cc160d94719b18e7f3b9c112628633aa6b832b2f09222b567f4b0dfe5b365)
BibTeX
@misc{dc03fe541febe2,
title = {Risk-Calibrated Proposal Transport for Finite-Particle Diffusion Steering — arXiv:2610.04171},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.04171},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_03fe541febe2; sha256 808cc160d94719b18e7f3b9c112628633aa6b832b2f09222b567f4b0dfe5b365},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.04171. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Inference-time steering combines pretrained diffusion experts or rewards without retraining by changing the dynamics that transport noise to data. Feynman-Kac correction compensates for proposal mismatch through importance-weighted sequential Monte Carlo (SMC), whose finite-particle behavior depends on the proposal. Variance-controlling guidance (VCG) improves that proposal by fitting a linear drift correction to minimize empirical log-weight-rate variance. Although its population optimum cannot worsen residual variance, finite-particle VCG can nearly eliminate its fitting residual while increasing residual risk on new states by orders of magnitude. The resulting update can degrade unweighted generation or accelerate particle collapse. We show that the centered Feynman-Kac rate is the normalized transport residual and that expected out-of-fit benefit is exactly population headroom minus coefficient-estimation penalty. Under regularity assumptions, a Wasserstein analysis bounds the unweighted proposal's terminal error using this residual. These results motivate Risk-Calibrated Proposal Transport (RCPT), which uses deletion leave-one-out residuals to calibrate the retained fraction of the VCG update, adding no model calls and only small linear-algebra overhead. Experiments on 2D checker distributions, scaffold decoration, molecular property optimization, and class-conditional CIFAR-10 generation demonstrate recovery from harmful fitted updates. Across molecular and image domains, RCPT mitigates harmful fitted updates and improves a broad range of terminal metrics relative to uncalibrated VCG. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Machine Learning (stat.ML) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.04171 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-06T05:24:05.807Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).