AI Briefing#reinforcement-learning
The latest AI and developer news about #reinforcement-learning, with the original source and a short summary.
The latest AI and developer news about #reinforcement-learning, with the original source and a short summary.
Mercor and SkyRL have released the complete RL training process for a 397B-parameter model's knowledge work agent.
Google Cloud AI Research and others proposed 'EnvHarness', a wrapper layer that dynamically reconstructs static benchmark environments to target agent weaknesses.
The ReViSQL-K2.6 model achieved human-level performance on Text-to-SQL benchmarks without scaffolding through RLVR and reward shaping.
CUA-Lite has released open-source tools for running, evaluating, and training computer use agents with local models.
Pollen Robotics and Hugging Face released Microduck, an open-source bipedal robot equipped with reinforcement learning software.
Hugging Face has released MicroDuck, a 10-inch open-source bipedal robot equipped with 15 actuators and various sensors.
Applied Compute has released AC2, an AI model training and deployment platform previously used by its internal research team.
IBM released three Granite 4.2 reasoning LLMs featuring 512K context and agentic RL.
IBM has released three Granite 4.2 reasoning LLMs featuring 512K context and agentic RL.
Agent harness optimization, inference-time structural evolution, and deep safety reassessment define this week's AI/ML paper trends.
Google published a technique in Nature that uses reinforcement learning to correct quantum computer errors in real time.
TaoLive unveiled the HAT technique to train small LLMs for real-time live commerce to adapt to evolving harnesses.
Ornith-1.5 has been released as a series of open models ranging from 397B to 9B parameters.
Agent Lightning v1.0 has been released as an RL training framework leveraging agent harnesses.
Ornith-1.5 achieves open-source SOTA through self-improvement via self-generated tasks and reinforcement learning
An open-source 30B model surpassed GPT-5.5 on the VibeWorlding benchmark, setting a new standard for 3D world generation.
As cybersecurity threat capabilities grow, OpenAI is adjusting its model development pace and pushing for enhanced security and alignment.
Tencent has released UI-Mate-27B, an open-source GUI agent that performs desktop tasks using natural language instructions and demo-based learning.
The reinforcement learning-based CUDA Agent has demonstrated superior CUDA kernel generation performance compared to existing models.
AReno, a toolkit supporting RL, SFT/DPO training, and agent RL for LLMs on a single node, has been released.
MIT and Harvard researchers proposed the 'Role Anchor' technique to address role drift in multi-LLM pipelines.
A large-scale analysis of reasoning performance and cross-lingual transfer effects in non-English and multilingual environments using GRPO.
Google is reportedly collaborating with AMD to design a next-generation TPU featuring integrated CPU cores to enhance reinforcement learning performance.
Microsoft AI introduced MAI-Code-1.1-Flash to GitHub Copilot, improving performance and speed while reducing costs to a quarter.