Trendoday

RL Post-Training in Diffusion Transformers Is Activation Steering

Reinforcement learning (RL) post-training substantially improves the visual quality, text rendering, and compositional accuracy of diffusion transformers. Because capabilities such as rendering reques

  • Upvotes8
#14Rank now –
#14Peak rank
3Days on list
274Score · 7d
Nov 20, 2026First seen
Rank · 7 days

About this paper

A fixed set of steering vectors added to a frozen diffusion transformer recovers most of the quality gains from reinforcement learning post-training, indicating that much of the improvement comes from a constant shift rather than prompt-specific transformations.

Reinforcement learning (RL) post-training substantially improves the visual quality, text rendering, and compositional accuracy of diffusion transformers. Because capabilities such as rendering requested text or composing multiple objects are inherently prompt-dependent, a natural hypothesis is that improving them requires learning complex, input-dependent transformations. To isolate how much of these gains an input-independent shift explains, we probe each RL-tuned model with a deliberately simple parameterization: a fixed set of steering vectors added to the block outputs of the frozen base model. On SD3.5-M, these vectors hold 65x fewer parameters than the RL-tuned LoRA (0.29M against…

reinforcement learning post-trainingdiffusion transformerssteering vectorsblock outputsfrozen base modelSD3.5-MRL-tuned LoRAvelocity predictionsRL-tuned checkpointspolicy-gradient RL
Upvotes
8
Authors
6
Published
Nov 19, 2026

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

Trending history

RL Post-Training in Diffusion Transformers Is Activation Steering was first seen on Hugging Face · Daily papers on Nov 20, 2026. Since then it has appeared in 23 checks on 3 different days. Its best position was #14 of 30, reached on Nov 22, 03:29 UTC. It is #14 right now, holding steady at #14 over the last 6 checks.

Rank history

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

#1#11#20#30 19:2906:2917:2904:2915:29 Nov 20, 19:29 UTC · #23Nov 20, 21:29 UTC · #16Nov 20, 23:29 UTC · #16Nov 21, 01:29 UTC · #16Nov 21, 03:29 UTC · #16Nov 21, 05:29 UTC · #16Nov 21, 07:29 UTC · #16Nov 21, 09:29 UTC · #15Nov 21, 11:29 UTC · #15Nov 21, 13:29 UTC · #15Nov 21, 15:29 UTC · #15Nov 21, 17:29 UTC · #15Nov 21, 19:29 UTC · #15Nov 21, 21:29 UTC · #15Nov 21, 23:29 UTC · #15Nov 22, 01:29 UTC · #15Nov 22, 03:29 UTC · #14Nov 22, 05:29 UTC · #14Nov 22, 07:29 UTC · #14Nov 22, 09:29 UTC · #14Nov 22, 11:29 UTC · #14Nov 22, 13:29 UTC · #14Nov 22, 15:29 UTC · #14

Last 14 days

Best rank per day (UTC).

Nov 9today

Metric history

▲ as reported by Hugging Face.

071320 19:2906:2917:2904:2915:29 Nov 20, 19:29 UTC · 8Nov 20, 21:29 UTC · 14Nov 20, 23:29 UTC · 14Nov 21, 01:29 UTC · 14Nov 21, 03:29 UTC · 14Nov 21, 05:29 UTC · 14Nov 21, 07:29 UTC · 14Nov 21, 09:29 UTC · 16Nov 21, 11:29 UTC · 16Nov 21, 13:29 UTC · 16Nov 21, 15:29 UTC · 17Nov 21, 17:29 UTC · 17Nov 21, 19:29 UTC · 17Nov 21, 21:29 UTC · 17Nov 21, 23:29 UTC · 17Nov 22, 01:29 UTC · 17Nov 22, 03:29 UTC · 18Nov 22, 05:29 UTC · 18Nov 22, 07:29 UTC · 19Nov 22, 09:29 UTC · 19Nov 22, 11:29 UTC · 19Nov 22, 13:29 UTC · 19Nov 22, 15:29 UTC · 20

Questions

Is RL Post-Training in Diffusion Transformers Is Activation Steering trending on HF Daily Papers right now?

Yes. As of Nov 22, 15:29 UTC it is #14, holding steady at #14 over the last 6 checks.

What is the highest rank RL Post-Training in Diffusion Transformers Is Activation Steering reached?

#14 on Hugging Face · Daily papers, on Nov 22, 03:29 UTC. Trendoday checks this list every 120 minutes.

How long has RL Post-Training in Diffusion Transformers Is Activation Steering been trending?

It was first seen on Nov 20, 2026 and has been on the list on 3 days (23 checks) since.