Shortcuts & Boundaries: RL Misalignment and Contextual Safety
How agentic RL can improve real work while learning contextual violations, and how verifiable feedback supports useful, transferable safety.
Read storyXinjie Shen
Blog
Notes on research, agents, and how the work gets made.
A Chat History of Making a Research Figure
The VHD-Play teaser figure, from the first draft to the final version.
Read story
What If the Bug Had a Dial?
FLAS turns model behavior into a continuous dial, from fixing creature-metaphor bugs to controlled brand voice, all at inference time without retraining.
Read story
New Preprint: CKA-Agent Achieves 96-99% Jailbreak Success on GPT, Gemini & Claude
Our new paper introduces CKA-Agent, a novel framework that bypasses commercial LLM guardrails through adaptive tree search over correlated knowledge.
Read story
