Add ClawBench to Datasets / Benchmarks

[reacher-z]

Section fit

ClawBench is a GUI agent benchmark covering everyday web tasks on live production websites, which aligns with the Datasets / Benchmarks section. It is appended chronologically after the most recent entry (June 2025).

What ClawBench is

• 153 everyday web tasks on 144 live production websites across 15 life categories
• 7 frontier models evaluated; the best passes 33.3%
• Unique mechanism: a submission-interception layer (Chrome extension + CDP) blocks only the final write request, letting agents run end-to-end on live sites without real-world side effects

Public references

• Paper: https://arxiv.org/abs/2604.08523
• Repo: https://github.com/reacher-z/ClawBench
• Site: https://claw-bench.com

Entry format

Follows the existing section convention: + [Title](paper link) (Month. Year) with Star / arXiv / Website shields beneath.

Affiliation disclosure

I am affiliated with the ClawBench project.

[reacher-z]

Checking in on this one — if the format's off for your list or you'd prefer a different section, I'll happily push a fix. No rush on my end.

[reacher-z]

Closing this earlier duplicate in favor of the newer PR #64, which supersedes it.