Jinpeng Wang

UI-MOPD achieves a remarkable balance between retaining existing capabilities and adapting to new platforms, with task success rates that challenge conventional approaches in GUI agent learning.

Attention-guided denoising can dramatically enhance reasoning performance in diffusion language models, outperforming traditional post-training methods.

Key contribution not extracted.

Unlock "white-box" reasoning in vision-language models: SegCompass's sparse autoencoder creates an interpretable bridge between visual perception and chain-of-thought, outperforming black-box alignment methods.

Spatial awareness is the secret ingredient to unlocking better visual in-context learning, boosting performance across diverse vision tasks.

Achieve state-of-the-art long-horizon video understanding by compressing multimodal memories into high-level semantic schemas, enabling efficient reasoning without losing crucial details.

Task-oriented dialogue agents can now learn to balance user satisfaction and operational costs, thanks to a new RL framework that optimizes for both.

Context-augmented RL lets smaller MLLMs punch *way* above their weight, rivaling much larger models on reasoning tasks while dodging reward hacking.

LLMs learn to recommend better by looking inside themselves, using intermediate layer activations to generate harder negatives on the fly.