Environment Duels benchmark scores LLMs as task authors and solvers
Across nine frontier models, the best solvers are not the best authors, and smaller models write the hardest environments.
Stories per day · 30 days
2 RSS
Relatedagentic-evaluation1embodied-ai1hm3d1llm1llm-benchmarks1model-evaluation1navigation1policy-adaptation1procthor1reinforcement-learning1
3 stories, newest first · one card per story
Stories tagged #environment-generation
From Failures to Learnable Worlds: Policy-Conditioned Environment Generation for Navigation Agents
WEDGE generates navigation training environments from a policy's validated failure modes, beating random and hard-example baselines.
Verifiable Hidden Dynamics Play: Generating Agentic RL Environments from Solved Mechanisms
VHD-Play builds 3,300 agentic RL environments and lifts Qwen3.6-35B-A3B's mean score from 0.204 to 0.815.