# Yonatan Bisk / CLAW @ CMU

> Connecting Language to Action and the World -- Yonatan Bisk's lab at Carnegie Mellon University, working on embodied AI, grounded language, robotics, and multimodal machine learning.

## Pages

- [Homepage](https://talkingtorobots.com/): About the CLAW lab and its research themes.
- [Publications](https://talkingtorobots.com/publications.html): Full, searchable list of papers.
- [Group](https://talkingtorobots.com/CLAW/): Current students, postdocs, and alumni.
- [Yonatan Bisk](https://talkingtorobots.com/yonatanbisk.html): PI homepage, bio, and contact FAQ.
- [Teaching](https://talkingtorobots.com/teaching.html): Courses taught at CMU.
- [Resources](https://talkingtorobots.com/resources.html): Benchmarks and code released by the lab.
- [Grad School FAQ](https://talkingtorobots.com/FAQ.html): Advice for prospective students.

## Recent Publications

- [Contextual Observer Grounding: Evaluating Situated Spatial Reasoning in Vision-Language Models](https://arxiv.org/abs/2609.06880): Findings of The 2026 Conference on Empirical Methods in Natural Language Processing (2026)
- [Amplified Does Not Mean Predictive: Reasoning Behaviors in Thinking Models](https://arxiv.org/abs/2608.13760): Third Conference on Language Modeling (2026)
- [Inductive Generalization for Robotic Manipulation](https://arxiv.org/abs/2606.20999): ArXiv (2026)
- [Lost in Reconstruction: Aligning Action Representations with Language in Vision-Language-Action Models](https://arxiv.org/abs/2608.10484): ArXiv (2026)
- [Flatness Preserves Instruction Following in Vision-Language-Action Models](https://arxiv.org/abs/2606.23641): ArXiv (2026)
- [ELASTIC: Efficiently Learning to Adaptively Scale Test-Time Compute for Generative Control Policies](https://arxiv.org/abs/2606.31132): ArXiv (2026)
- [IndoorR2X: Indoor Robot-to-Everything Coordination with LLM-Driven Planning](https://arxiv.org/abs/2603.20182): IEEE/RSJ International Conference on Intelligent Robots and Systems (2026)
- [When Should AI Read the Room? Public Perceptions of Social Intelligence in AI Agents](https://arxiv.org/abs/2605.29938): ArXiv (2026)
- [Position: Evaluation of {ML} Resource Utilization Requires Model Life Cycle Assessment](https://openreview.net/forum?id=8voi61IiZC): Forty-third International Conference on Machine Learning Position Paper Track (2026)
- [MM-SeR: Multimodal Self-Refinement for Lightweight Image Captioning](https://arxiv.org/abs/2508.21451): Computer Vision and Pattern Recognition Conference (2026)