← Search

Kshitij Sachan

2 accepted papers

2024

AI Control: Improving Safety Despite Intentional Subversion

ICML 2024oral

As large language models (LLMs) become more powerful and are deployed more autonomously, it will be increasingly important to prevent them from causing harmful outcomes. To do so, safety measures either aim at making LLMs try to avoid harmful outcomes or aim at preventing LLMs from causing harmful o…

2024

Debating with More Persuasive LLMs Leads to More Truthful Answers

ICML 2024oral

Common methods for aligning large language models (LLMs) with desired behaviour heavily rely on human-labelled data. However, as models grow increasingly sophisticated, they will surpass human expertise, and the role of human evaluation will evolve into non-experts overseeing experts. In anticipatio…