Jie Song

LLM agents can substantially improve their task-solving abilities by treating skills as long-lived, experience-aware, and testable assets within a managed lifecycle.

Pruning detrimental LoRA modules can lead to substantial performance gains in multi-task models, challenging the assumption that all components contribute positively.

Achieve faithful textile pattern generation by disentangling clothing features and guiding a diffusion model with fine-grained alignment, outperforming existing image-to-image methods.

Achieve diffusion-level perceptual quality in monocular depth estimation at 40x the speed, by replacing the slow initial diffusion steps with a fast ViT-based depth map and refining in a compact latent space.

DriveFix tackles the "shaky camera" problem in 4D driving scene reconstruction, producing significantly more stable and coherent novel views by explicitly modeling spatio-temporal dependencies.

Ditch the min-max: Fuz-RL offers a fuzzy-measure guided approach to safe RL that achieves distributional robustness without complex optimization.

VLMs still can't reason about spatial logic in real-world scenes, but a new benchmark and scene graph method shows how to make progress.