Debating with More Persuasive LLMs (ICML 2024 Best Paper)

Debating with More Persuasive LLMs (ICML 2024 Best Paper)

Why it matters

The first empirical evidence that debate-based oversight actually raises the accuracy of non-expert judges

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2402.06782

Added in Week 35, 2026 · See that week

Same week

The Unreasonable Effectiveness of Eccentric Automatic Prompts publishedPaperSelf-Discover publishedPaperDeepSeek introduces GRPO (DeepSeekMath)Paper

Nearby on the same axis

Sleeper AgentsPaperWeak-to-strong generalizationOfficialOpenAI publishes the first Model SpecOfficialFrontier Safety Framework v1Official