Guanghui Wang

Generated translation references outperform ground truth by 8.65 CEA100 points, showcasing a novel approach to literary translation challenges.

LLMs can teach themselves new tricks: a simple self-improvement loop, "Ratchet," lets a frozen LLM agent significantly boost its coding performance by autonomously managing its own library of natural-language skills.

Vision Mamba's accuracy can be significantly improved just by swapping out the default discretization method, with bilinear discretization offering the best balance of performance and efficiency.

End-to-end prompt optimization is often a waste of time and money, succeeding only when coaxing models into specific output formats they're already capable of.

Expert-written rules for coding agents are often useless or even harmful, with random constraints working just as well and negative constraints outperforming positive directives.