Ye Chen

Sampling rollouts via GRPO actively degrades tool-selection training as policies concentrate, whereas exact policy optimization over enumerable combinatorial tool spaces eliminates vanishing gradients and boosts accuracy by up to 14.2 points.

Adaptive similarity margins in HN-CLIP boost retrieval accuracy by up to 4.3% while training 2.4x faster than leading methods.

Unifying motion and camera controls in a single visual representation leads to unprecedented improvements in video generation fidelity and robustness.

Models may ace aesthetics but falter on geometry—SVGEval reveals the critical gaps in text-to-SVG generation evaluation.

Incorporating visual feedback into prompt optimization leads to significant performance improvements in vision-language tasks, revealing previously unrecognized error patterns.

TARA redefines prompt optimization by routing specific failures to tailored repair operators, achieving superior semantic accuracy without the need for generator retraining.

Transforming video generation from a pixel sampling problem to a structured orchestration of the physical world, WNM enables unprecedented control and efficiency in content creation.

LLMs show significant vulnerability to logical fallacies, with distinct profiles of resilience that could inform future model training strategies.

Ling-2.6 and Ring-2.6 achieve unprecedented efficiency in agentic intelligence, enabling instant responses and deep reasoning at trillion-parameter scale.

You can slash LLM prompt evaluation costs by 35-60% without sacrificing accuracy by intelligently selecting which examples to use.