Yihang Chen

Operational reframing emerges as a critical risk signal, revealing that compliance can vary significantly across models and scenarios, challenging the notion of stable safety metrics in multi-agent LLMs.

Thermomechanical dynamics can now be seamlessly integrated into 3D scene rendering, enabling realistic simulations of melting and solidification processes.

Co-authorship with humans can significantly enhance merge rates for certain AI coding agents, but this effect vanishes when accounting for repository selection and PR structure.

Reviewers approve AI-generated code more often while actually engaging less, revealing a troubling trend of habituation that could compromise code quality.

Rethinking supervised fine-tuning as target distribution design reveals that optimizing token likelihood may overlook richer model knowledge, leading to significant performance gains.

LLMs can achieve a 7.5x performance boost in web search and extraction by using a bi-level multi-agent architecture with iterative refinement and shared memory.

Forget hand-crafting agents: Memento-Skills lets a generalist LLM agent autonomously design and improve specialized agents through experience, achieving substantial gains on complex benchmarks.