Jiaqi Wang

DOPD reveals that intelligently routing supervision based on advantage gaps can significantly enhance capability transfer in distillation, outperforming conventional methods.

Forgetting earlier observations, not decision-making flaws, is the primary source of errors in multimodal LLMs navigating complex tasks.

PermaVid achieves unprecedented long-term consistency in video generation, even after significant edits, by intelligently disentangling appearance and geometry in memory.

Human raters overwhelmingly prefer JoyAI-VL-Interaction over existing video-call assistants, showcasing a leap in real-time interaction capabilities.

CapRL++ redefines caption quality through utility, enabling models to produce high-fidelity descriptions without the constraints of traditional supervised fine-tuning.

Light-WAM achieves high-performance robot manipulation with just 0.44B parameters, revolutionizing the efficiency of World Action Models.

By co-evolving experts through bidirectional policy distillation, CoPD achieves all-in-one integration of text, image, and video reasoning, outperforming domain-specific experts and suggesting a new training paradigm.

Forget external teachers – the best way to boost your RL model's performance is to learn from its future self.

Self-distillation in LLMs can leak information and destabilize training, but combining it with verifiable rewards yields a sweet spot for improved convergence and stability.

Robot control gets a whole lot faster: ProbeFlow slashes action decoding latency by 14.8x in Vision-Language-Action models, all without retraining.