README: describe both leaderboards (dataset page for open-weight models, Space for all models)
#7
by ajayjagan2511 - opened
Now that open-weight results are landing on this dataset's leaderboard (via .eval_results PRs on the DeepSeek, Hy4, Mistral and GLM model repos), the README no longer says we use the Space instead of the built-in leaderboard. It describes both:
- Open-weight models: the leaderboard on this dataset page, which updates on its own; add a model by opening a PR on its repo with a results file under
.eval_results/(docs). - All evaluated models, including API-only ones (Gemini, Claude, GPT, Grok, Qwen Max), which have no model repo: the CWE-bench leaderboard Space; add a model via a PR on the Space or cyber@collinear.ai.
Only the Leaderboard section's text changes; the results table and the rl-environment tag are unchanged.
chalo2000 changed pull request status to merged
