#grpo
2 briefs tagged #grpo.
Research
ClawGym II targets black-box RL for agent harnesses
The arXiv paper proposes training agents through opaque, long-horizon harnesses.
HF Daily Papers+1 outlet·1d ago
Research
Video-DeepResearch targets video-grounded AI agents
The arXiv paper introduces a benchmark and 35B-A3B model for multi-hop video VQA with web exploration.
HF Daily Papers+1 outlet·Aug 4