Yuan-Chen Chang
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
CAT-alyzing Honesty: Continuous Adversarial Training for Transferable LLM Honesty
According to the abstract excerpt, the authors introduce CATCH, which trains a model toward honest answers and away from dishonest ones using only hundreds of targets. They report that it generalizes across benchmarks and beats prompting and supervised fine-tuning, tested on Llama-3.3-70B-Instruct and a Qwen model.