Agent World Model in the Wild

Computer-Use Agent Environment Synthesis

WildAWM-4B is the WildAWM agent model, a Qwen3.5-4B checkpoint post-trained on the WildAWM XAI task set. It is served text-only with a 262144-token context, and it keeps the Qwen tool-calling and reasoning formats used by the evaluation harness.

🔮 Resources

Resource	Link
🌐 Project page	WildAWM
💻 Code (synthesis, evaluation)	WildAWM repository
📦 Task set	ChilleD/WildAWM
🤖 WildAWM-35B-A3B	ChilleD/WildAWM-35B-A3B

📦 Setup

uv pip install "sglang==0.5.15.post1"

🔥 Serving

This is the simplest single-GPU form of the launch used by the WildAWM pipeline:

python -m sglang.launch_server \
  --model-path ChilleD/WildAWM-4B \
  --served-model-name wildawm-4b \
  --host 0.0.0.0 --port 30000 \
  --dtype bfloat16 \
  --context-length 262144 \
  --mem-fraction-static 0.85 \
  --tool-call-parser qwen3_coder \
  --reasoning-parser qwen3

Please find more details in WildAWM repository for how to use it for evaluation.

Citation

If you find this work useful, please cite:

@article{wang2026agentworldmodelwild,
  title={Agent World Model in the Wild: Computer-Use Agent Environment Synthesis},
  author={Zhaoyang Wang and Qianhui Wu and Haorui Wang and Simon Yu and Baolin Peng and Canwen Xu and Hao Cheng and Wenlin Yao and Boyi Liu and Yite Wang and Jianfeng Gao and Huaxiu Yao},
  year={2026},
  eprint={2610.26828},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2610.26828},
}