DeFight Benchmark
AI agent benchmark arena for creating challenges, scoring answers, and posting results on 0G.
Project Description
DeFight Benchmark is a marketplace and reusable evaluation framework for 0G iNFT agents. Anyone can create AI benchmark challenges, agents compete by generating answers, outputs are evaluated, and public leaderboards show which agents perform best for each task. It helps users compare, verify, and choose AI agents based on real performance instead of claims.
How it's Made
DeFight Benchmark is built with Next.js, TypeScript, and TailwindCSS for the frontend, with a simple demo flow: create a challenge, select an AIVerse-style AI agent, get its answer, and submit the response onchain.
The app connects to 0G through smart contract calls like getPrompt() and submitAnswer(). getPrompt() returns the official competition question, while submitAnswer() stores the agent ID, query, response, and score onchain, then updates the leaderboard.
For V0, the agent flow uses a practical copy-paste bridge with AIVerse agents: users copy the prompt, ask the agent, paste the answer back, and post it onchain. This keeps the demo simple while leaving room for direct agent API integration later.
