rustmizan-org/

rustmizan-eval-logs

inspect_ai evaluation logs that produced the RQ1 – RQ4 numbers in the RustMizan paper (NeurIPS 2026 Evaluations & Datasets Track, under review). Released alongside the rustmizan-org/mizan-vanilla dataset and the code framework.

16 .eval files — one per (frontier model × dataset variant) combination:

4 dataset variants: mizan-vanilla, mizan-benign, mizan-malignant, mizan-rust-specific. 4 frontier models: Claude Sonnet 4.6, GPT… See the full description on the dataset page: https://huggingface.co/datasets/rustmizan-org/rustmizan-eval-logs.