Crimson AI NewsA CrimsonLingua Network service
EN ع
← Back to news
Research paper Hugging Face

Predictive Divergence Masks: A New Direction for LLM Reinforcement Learning

AI By Crimson AI Hugging Face Papers 26 July 2026 · 00:00 13 views
Share: X Telegram

Researchers propose predictive divergence masks to replace the ratio-based direction criterion in PPO-style RL for LLMs, improving alignment between policy updates and divergence constraints.

Predictive Divergence Masks: A New Direction for LLM Reinforcement Learning

Key points

Reinforcement learning (RL) for large language models (LLMs) often relies on trust-region masks to stabilize off-policy updates. The dominant PPO-style approach uses the sampled-token importance ratio for two criteria: a proximity criterion (whether the policy has moved too far from the behavior policy) and a direction criterion (whether the update pushes it farther away). Recent work, DPPO, improved the proximity criterion by using a probability divergence instead of a ratio-based test, but retained PPO's direction criterion.

The authors observe that the ratio-based direction criterion is a single-sample proxy that can disagree in sign with the change of the divergence defining the proximity criterion. To address this, they propose the predictive divergence mask, which asks whether the next policy-gradient step will increase or decrease the same divergence used by the trust region. For discrete softmax policies, they derive this prediction in closed form.

Because production rollout engines expose only a truncated (top-K) view of the vocabulary, the authors develop two lightweight top-K estimators for this prediction. Detailed analysis shows the divergence-based direction is better aligned with the realized change of the divergence than the sampled ratio, and the resulting masks improve RL training across model scales and precision settings.

Source
Hugging Face · Hugging Face Papers
Related news
Research paper
Hugging Face 31 Aug 2026

Hugging Face Unveils StepGuard: Step-Level Guardrails for Safer AI Agents

StepGuard, a new step-level guard model from Hugging Face, audits agent actions before execution, reducing attack success rates by...

1
Research paper
Hugging Face 31 Aug 2026

Hugging Face Researchers Unveil ABot-Recon for Stable Long-Horizon 3D Reconstruction

ABot-Recon, a new streaming 3D reconstruction model from Hugging Face, achieves stable long-horizon performance using only local t...

1
Research paper
Hugging Face 31 Aug 2026

ContextPilot: Teaching Agents Proactive Context Management via Fine-Grained RL

Hugging Face researchers introduce ContextPilot, a framework that enhances long-horizon agent reasoning by expanding context-editi...

1