The proposed thesis, “Chance Prediction under Distribution Shift: Diagnosis, Calibration, and Guarantees for Visual Autonomy and Language Agents”, outlines a vision for detecting, adapting, and calibrating confidence under 4 types of shifts: observation, dynamics, initial state, and controller.
The proposal builds on a significant list of papers where Zhenjiang contributed:
- Zhenjiang Mao*, Jiawen Wu*, Gabriel Wagner, and Ivan Ruchkin.
Anomaly-Informed Confidence Calibration for Vision-Based Safety Prediction [Arxiv] [Video].
In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Pittsburgh, Pennsylvania, 2026. * Co-first authors. - Zhenjiang Mao*, Anirudhh Venkat*, Artem Bisliouk, Akshat Kothiyal, Sindhura Kumbakonam Subramanian, Saithej Singhu, Ivan Ruchkin.
Confidence over Time: Confidence Calibration with Temporal Logic for Large Language Model Reasoning [Arxiv].
In the Findings of the Association for Computational Linguistics (ACL), San Diego, California, 2026. * Co-first authors. - Zhenjiang Mao, Mrinall Eashaan Umasudhan, Ivan Ruchkin.
Physically Interpretable World Models via Weakly Supervised Representation Learning [Arxiv] [IEEE] [Slides] [Poster] [Demo] [Github].
In Proceedings of the 17th ACM/IEEE International Conference on Cyber-Physical Systems (ICCPS), Saint Malo, France, 2026. - Carson Sobolewski, Zhenjiang Mao, Kshitij Vejre, Ivan Ruchkin.
Generalizable Image Repair for Robust Visual Control [Arxiv] [Poster 1] [Poster 2] [Github] [Video] [Slides].
In Proceedings of the International Conference on Intelligent Robots and Systems (IROS), Hangzhou, China, 2025. - Jordan Peper*, Zhenjiang Mao*, Yuang Geng, Siyuan Pan, Ivan Ruchkin.
Four Principles for Physically Interpretable World Models [Arxiv] [OpenReview] [Github] [Poster] [Slides].
In Proceedings of the 2nd International Conference on Neuro-symbolic Systems (NeuS), Philadelphia, PA, 2025. * Co-first authors. - Zhenjiang Mao, Carson Sobolewski, Ivan Ruchkin.
How Safe Am I Given What I See? Calibrated Prediction of Safety Chances for Image-Controlled Autonomy [PMLR] [Arxiv] [Poster 2023] [Poster 2024] [Github].
In Proceedings of the Annual Learning for Dynamics & Control Conference (L4DC), Oxford, UK, 2024. - Zhenjiang Mao, Mrinall Eashaan Umasudhan, Ivan Ruchkin.
How Safe Will I Be Given What I Saw? Calibrated Prediction of Safety Chances for Image-Controlled Autonomy [Arxiv] [Github].
In submission, 2025.