razzant/ouroboros-benchmark-explorer

Ouroboros Benchmark Explorer

This static Space reads small result files from pinned commits in four public Ouroboros benchmark datasets. It has no backend and asks for no API key.

The OSWorld view lets you filter all 361 tasks, inspect the official reward, read the original prompt and the final or feasibility record, and open every published source file. CL-Bench and SWE-bench Pro link to their full trace packages and reproduction notes.

• —Paper: <https://huggingface.co/papers/2608.08311>
• —Source: <https://github.com/razzant/ouroboros>
• —Benchmark evidence: <https://ouroboros-agent.ai/benchmarks/>
• —Collection: <https://huggingface.co/collections/razzant/ouroboros-benchmark-evidence-6a6e7fec1cfdc2d8d93ed3a3>