Hey! I'm a M.S. grad student researcher currently working on my thesis under Prof. Lily Weng. Previously, I completed my BS in Computer Science Honors an...
TLDR: NLAs are a recent black box mech interp method for verbalizing model internals. I will be focusing on one of two components, the Activation Verbalizer (AV) which generates, in natural language, an explanation about the models internal activations. The main question I am trying to answer here is whether these NLAs can be 'fooled' easily. I ran two small stress tests on the AV. First, I prefix…
TLDR: So there has been recent discourse on 𝕏 , and recent news of major cyber attacks that were done with the help of AI. The missing frame here is the dual-use gap : as AI models become more capable, they create more upside for defenders and more downside for attackers. The gap between the benefits and the harmful effects is getting wider. I know that sounds obvious, but I think people are…
TLDR : Pragmatic interp sounds great in the sense that you get to keep interp tools while actually moving safety metrics, but looking a bit closer it's kinda a trap. You pay interp's overhead but get judged against black-box baselines that don't, so the work that survives is whatever cleared that bar, not whatever produced understanding. The two scoreboards (understanding vs intervention) don't…
This post summarizes work from my project in CSE-291A: Systems for LLMs and AI Agents at UC San Diego, under the guidance of Professor Yiying Zhang . The Problem LLM agents work in a loop: reason about what to do, call an external tool (web search, file reader, calculator, etc.), observe the result, repeat. The bottleneck isn't where you'd expect. Tool calls finish in milliseconds, but the model…
TLDR : Moltbook is interesting in the sense that we have thousands of AI agents in one social environment, but looking a bit closer fundamentally it is mid. It's just a mirror of human prompters, not actual emergent AI behavior. Identical prompts/contexts would probably cause the agents to converge to an equilibrium point - THAT would be interesting to see at that scale what that would actually…