[NeurIPS 2025 Spotlight] Reasoning Environments for Reinforcement Learning with Verifiable Rewards

parallel opinions → anonymous peer review → synthesis. Supports Claude, GPT-5.4, Grok-4, Gemini.

Look around through a robot's eyes — 6-DOF stereo camera arm with WebXR VR teleop

Experimental adapter for fine-tuning Qwen3-VL as a Vision-Language-Action (VLA) model

Collection of LLM completions for reasoning-gym task datasets

System 2 Reasoning Link Collection

Building the cognitive-core to solve ARC-AGI-2

Minimal hackable GRPO implementation