Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
University of Illinois Urbana-Champaign
4
0
9
4
Coding agents fix 40%+ more bugs when they're shown whole files verbatim, not compressed snippets or similar-but-unnecessary code.
A single weight edit, requiring only forward passes, can permanently disable an LLM's ability to follow specific harmful instructions, even when those instructions are rephrased or embedded in new prompts.
Sparse autoencoders unlock a shared, interpretable coordinate system for matching symbols across diverse AI systems, outperforming dense baselines by up to 45 points.
A single global topology predictor can unlock substantial performance gains in Transformers, outperforming layer-wise routing and achieving better results with less data.