lechmazur/debate — GitHub trending stats & insights

Featured

lechmazur/debate

Adversarial multi-turn benchmark for LLM debate quality, using side-swapped matchups and multi-model judging to rank models by judged debate performance.

Visit GitHub

Social mentions

Recent discussions about this repository across the web

Each matchup is debated twice on the same motion with sides swapped, cancelling side bias. A three-model judge panel decides winner and margin. Fable 5 won the head-to-head, but Opus 5 won by wider…

@LechMazur · x.com

Each matchup runs twice on the same topic with sides swapped. A three-model judge panel then decides winner and margin. Fable 5 still hasn't lost a single side-swapped matchup aggregate debate.

@LechMazur · x.com

This benchmark tests knowledge under stress, counterargument handling, strategic coherence, and evidence discipline.

@LechMazur · x.com

Bradley-Terry is Elo-like. Scale is relative within this pool and centered near 1500. This is not an absolute capability score. Topics:

@LechMazur · x.com

Topics:

@LechMazur · x.com

Repository activities

repository's daily and monthly activities across stars, forks, merged PRs, issues, and closed issues