Benchmark source URL
https://github.com/TIGER-AI-Lab/ClawBench
Anything else? (who owns it, what the score means; optional)
ClawBench is maintained by the NAIL Group in the TIGER-AI-Lab repository. It evaluates 153 everyday tasks across 144 live websites and 15 categories, with isolated containers, final-request interception, and five evidence layers (session replay, screenshots, HTTP traffic, browser actions, and agent messages). Paper: https://arxiv.org/abs/2604.08523 Project: https://claw-bench.com/
We pull scores from your own published results, so a structured data
file (JSON/CSV) in the repo is ideal. A public results page works too.
Benchmark source URL
https://github.com/TIGER-AI-Lab/ClawBench
Anything else? (who owns it, what the score means; optional)
ClawBench is maintained by the NAIL Group in the TIGER-AI-Lab repository. It evaluates 153 everyday tasks across 144 live websites and 15 categories, with isolated containers, final-request interception, and five evidence layers (session replay, screenshots, HTTP traffic, browser actions, and agent messages). Paper: https://arxiv.org/abs/2604.08523 Project: https://claw-bench.com/
We pull scores from your own published results, so a structured data
file (JSON/CSV) in the repo is ideal. A public results page works too.