47 papers this week in Reinforcement learning12 active threadsbusiest: Agent Security & Attacks

Reinforcement learning research

Generative media

Persistent memory and reinforcement learning reduce drift in long driving simulations

WAM-Sim combines a persistent hierarchical memory (structural, scene, instance tokens) with an online flow-matching reinforcement learning framework to prevent structural drift, identity mutations, and quality decay in long-horizon driving simulations. On the nuScenes benchmark, it achieves FID=3.47 and strong downstream perception; on self-collected minute-scale data, it maintains high scene and identity coherence.

2 days ago
Reinforcement learningAMZN ▼0.3%

User-agnostic reward component, not the personalized one, narrows AI recommender picks

The authors trained LLM recommenders with collaborative-filtering scores as reinforcement learning rewards and found that directly using these scores made the policy concentrate on far fewer items, dropping distinct top-1 picks by 36-80%. By decomposing each score into a user-independent item mean and a user-specific residual, they showed the item-mean component drives the collapse. Simply subtracting or standardizing it restored diversity: distinct top-1 items rose by up to 9.8 times with comparable ranking accuracy.

2 days ago
Efficiency & systems

New system halves compute waste during AI response generation training

Researchers developed FlexFloor, a system that cuts wasted compute time in reinforcement-learning post-training for mixture-of-experts language models. By releasing workers that have finished generating their responses while keeping their experts available for ongoing responses, FlexFloor boosts response tokens per occupied compute time by 65% and overall throughput by 53% with no loss in output quality.

2 days ago