Agentic Critical Training
An agent-training framework that converts critique and revision into a learning signal for more reliable multi-step behavior.
AI Agents
Turn models into purposeful systems. We develop agents that couple reasoning with action, critique, experimentation, and safety evaluation—especially where success requires long-horizon coordination.
Research question
How can an agent decide what to do, verify its progress, and recover when the first plan fails?
We develop agents that couple reasoning with action, critique, experimentation, and safety evaluation—especially where success requires long-horizon coordination.
- Planning and verification
- Tool use
- Self-critique
- Agentic evaluation
Representative projects
Turn models into purposeful systems.
Each project connects its central idea with papers, code, datasets, demonstrations, and public explanations.
Selected publications
The papers behind the projects.
Browse the full publication database for the broader body of work in reasoning control.
Agentic Critical Training
BibTeX
@misc{liu2026agenticb95b,
title = {Agentic Critical Training},
author = {Weize Liu and Minghui Liu and Sy-Tuyen Ho and Souradip Chakraborty and Xiyao Wang and Furong Huang},
year = {2026},
eprint = {2603.08706},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2603.08706},
}Imagine, Verify, Execute: Agentic Exploration with Vision-Language Models
9th Annual Conference on Robot Learning (CoRL), 2025
BibTeX
@inproceedings{lee2025imagineef14,
title = {Imagine, Verify, Execute: Agentic Exploration with Vision-Language Models},
author = {Seungjae^ Lee and Daniel Ekpo^ and Haowen Liu and Furong Huang and Abhinav Shrivastava and Jia-Bin Huang},
booktitle = {9th Annual Conference on Robot Learning (CoRL), 2025},
year = {2025},
eprint = {2505.07815},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2505.07815},
}PropensityBench: Evaluating Latent Safety Risks in Large Language Models via an Agentic Approach
The Fourteenth International Conference on Learning Representations (ICLR), 2026
BibTeX
@inproceedings{sehwag2026propensitybench0e64,
title = {PropensityBench: Evaluating Latent Safety Risks in Large Language Models via an Agentic Approach},
author = {Udari Madhushani Sehwag and Shayan Shabihi and Alex McAvoy and Vikash Sehwag and Yuancheng Xu and Dalton Towers and Furong Huang},
booktitle = {The Fourteenth International Conference on Learning Representations (ICLR), 2026},
year = {2026},
eprint = {2511.20703},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2511.20703},
}