Trevor Darrell

A unified generative model for 3D dynamics can predict, inpaint, and control complex SE(3) trajectories across diverse agents and scenes with a single network.

Predicted-state matching enables web agents to achieve superior action selection by ensuring that their state predictions are discriminative, leading to better task performance.

Models trained on CoVA-SFT outperform existing multimodal reasoning baselines by over 2x, revealing the potential of structured visual-textual integration in AI.

Leveraging weak supervision from clinical reports boosts CT scan segmentation accuracy by up to 22% with minimal labeled data.

Playful learning strategies enable robots to acquire skills that boost performance on new tasks by over 20%, transforming how we approach robotic skill acquisition.

NeRFs can revolutionize forest mapping by delivering high-fidelity 3D reconstructions that mitigate the flaws of traditional methods.

Extracting action signals from 32,041 hours of human video enables CAIP to outperform leading vision encoders in robotic manipulation tasks by over 30%.

Tactile-reactive policies can boost robotic manipulation success rates by over 30% through innovative data collection and a new Mixture-of-Transformers architecture.

Stateful Visual Encoders enable VLMs to leverage prior visual context, leading to substantial performance gains in multi-image tasks.

Achieve globally consistent 3D reconstruction over sequences exceeding 19,000 frames by combining test-time training with sliding window attention, outperforming prior state-of-the-art methods by over 74% on ATE on KITTI.