FightBench
github.com/aipdv/fightbench- Category
- Developer Tools
- Rank
- No. 2426Tools index
- Type
- TOOL
- Use case
- Model & Agent Evaluation
- Interfaces
- CLI
- Date
About
FightBench is a replay-verified benchmark for AI decision models. Its first edition has a model play Liu Kang in Mortal Kombat II (Genesis) against the 15 VeryHard CPU fights, 10 repeats each, reading game memory instead of pixels. Models are ranked by mean damage dealt, and a maintainer replays each submitted action log against the game to confirm the outcome. Model name, cost and latency are reported by the submitter and are not re-checked.
Tags
benchmarkllm-evalgamesmortal-kombatdecision-modelspython
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.