Jeffrey Ichnowski

Tests on five architecturally diverse VLAs suggest that adaptation cost in VLAs is structured enough to measure before fine-tuning begins, and introduces a pipeline that observes, diagnoses, allocates, and adapts.

ModAR is introduced, the first WAM to autoregressively denoise multiple future modalities before predicting actions, and it is found that ModAR's sequential generation outperforms existing WAM formulations, with the highest average success rate at all evaluated data scales.

HSR boosts robot manipulation success rates by over 21% by leveraging hierarchical skill retrieval, even with minimal task-specific data.

Robots can now autonomously learn from their failures, boosting success rates by over 17% without human intervention.

Joint image-depth generation can be achieved with a single model trained on sparse data, outperforming existing methods by a significant margin.

AHEAD transforms static VLA models into dynamic manipulators, achieving up to 97% success in scenarios where traditional methods fail spectacularly.

Forget expensive tactile sensors – microphones and a lightweight CNN can give you accurate, real-time in-hand slip estimation.

Soft robots can now mimic human manipulation skills more effectively by explicitly reasoning about contact forces, outperforming traditional kinematic imitation and learning-based methods.

Robots can now learn manipulation skills from ordinary human videos, thanks to a 3D point tracking method that bridges the embodiment gap and requires only 20 robot demonstrations.