Search papers, labs, and topics across Lattice.
School of Engineering, University of Edinburgh
3
0
6
STACE reveals that autonomous stress-testing with LLM agents can dramatically enhance the evaluation of concept erasure, outperforming traditional methods in robustness and adaptability.
VLMs struggle with emotion recognition not because they lack visual acuity, but because pre-training exacerbates dataset biases and sparse temporal sampling misses crucial micro-expressions.
Causal discovery's reliability hinges on often-overlooked sampling choices, as even state-of-the-art methods wobble significantly with varying rates and window lengths.
