Skytliang/Multi-Agents-Debate
MAD: The first work to explore Multi-Agent Debate with Large Language Models :D
# Technical Summary Implements a multi-agent framework where paired LLM agents debate opposing viewpoints to overcome "degeneration of thoughts"—the tendency of single-agent self-reflection to entrench biases and miss blind spots. The system designates agents as affirmative/negative sides with a judge arbiter, leveraging adversarial interaction to correct distorted reasoning and incorporate external feedback. Evaluated on Counterintuitive QA and Commonsense Machine Translation tasks using OpenAI's API with interactive CLI and Gradio demo support.
532 stars.
Stars
532
Forks
57
Language
Python
License
GPL-3.0
Category
Last pushed
Dec 16, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/Skytliang/Multi-Agents-Debate"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
betagouv/ComparIA
Open source LLM arena created by the French Government
liuxiaotong/ai-dataset-radar
Multi-source async competitive intelligence engine for AI training data ecosystems with...
llm-ring/lmring
Open-source, self-hostable LLM arena with model compare, voting, and leaderboards
Arnoldlarry15/ARES-Dashboard
AI Red Team Operations Console
khoren93/ai-debates
Orchestrate epic battles between 600+ AI models (GPT-5, Gemini 3, DeepSeek R1). Real-time...