Trendoday

FewStepRL: Fine-tuning Few-step Diffusion Models with Reinforcement Learning

Reinforcement learning (RL) has become a standard post-training method for diffusion models, but most RL objectives roll out the current policy at every iteration, and with a many-step sampler these r

  • Upvotes4
–Rank now
#15Peak rank
2Days on list
116Score · 7d
Nov 10, 2026First seen
Rank · 7 days

About this paper

Reinforcement learning objectives for diffusion models are reformulated for few-step samplers through x-space transformation, gradient decomposition, and anchor adaptation, yielding modular few-step variants of several methods and showing reward-gradient methods excel with differentiable rewards…

Reinforcement learning (RL) has become a standard post-training method for diffusion models, but most RL objectives roll out the current policy at every iteration, and with a many-step sampler these rollouts typically dominate the cost of training. Few-step models cut this cost, yet they model an average velocity over an interval, whereas existing RL objectives are built for an instantaneous velocity with a many-step sampler. We adapt RL objectives to few-step models systematically in three phases, namely a transformation, a decomposition and an adaptation. The transformation to x-space expresses each objective we consider through a common policy prediction, at the clean end by default and…

reinforcement learningdiffusion modelspost-trainingmany-step samplerfew-step modelsaverage velocityinstantaneous velocityx-spacepolicy predictionflow-map objectives
Upvotes
4
Authors
7
Published
Nov 5, 2026

Summary and abstract excerpt via Hugging Face Papers ↗ · arXiv.

Trending history

FewStepRL: Fine-tuning Few-step Diffusion Models with Reinforcement Learning was first seen on Hugging Face · Daily papers on Nov 10, 2026. Since then it has appeared in 10 checks on 2 different days. Its best position was #15 of 30, reached on Nov 11, 05:28 UTC. It dropped off the list after Nov 11, 07:28 UTC.

Rank history

Position on Hugging Face · Daily papers at every check (lower is better).

#1#11#20#30 13:2817:5822:2802:5807:28 Nov 10, 13:28 UTC · #28Nov 10, 15:28 UTC · #26Nov 10, 17:28 UTC · #22Nov 10, 19:28 UTC · #22Nov 10, 21:28 UTC · #18Nov 10, 23:28 UTC · #20Nov 11, 01:28 UTC · #21Nov 11, 03:28 UTC · #17Nov 11, 05:28 UTC · #15Nov 11, 07:28 UTC · #20

Last 14 days

Best rank per day (UTC).

Nov 3today

Metric history

Platform metric as reported by Hugging Face.

0358 13:2817:5822:2802:5807:28 Nov 10, 13:28 UTC · 4Nov 10, 15:28 UTC · 4Nov 10, 17:28 UTC · 5Nov 10, 19:28 UTC · 6Nov 10, 21:28 UTC · 7Nov 10, 23:28 UTC · 7Nov 11, 01:28 UTC · 7Nov 11, 03:28 UTC · 8Nov 11, 05:28 UTC · 8Nov 11, 07:28 UTC · 8

Questions

Is FewStepRL: Fine-tuning Few-step Diffusion Models with Reinforcement Learning trending on HF Daily Papers right now?

Not at the moment. It was last on the list on Nov 11, 07:28 UTC, after 2 days of trending.

What is the highest rank FewStepRL: Fine-tuning Few-step Diffusion Models with Reinforcement Learning reached?

#15 on Hugging Face · Daily papers, on Nov 11, 05:28 UTC. Trendoday checks this list every 120 minutes.

How long has FewStepRL: Fine-tuning Few-step Diffusion Models with Reinforcement Learning been trending?

It was first seen on Nov 10, 2026 and has been on the list on 2 days (10 checks) since.