anzenresearch/cybench-irt-pilot
Add author attribution
Standalone IRT artifact: add introduction and motivation, remove version history and archived revisions, fix LaTeX escaping and internal cross-reference links, correct distillation caption.
v3 re-audit: converged MAP fit with declared priors; report score censoring and item degeneracy; replace tautological calibration figure; exact permutation null for distillation; disclose target leakage; add dispersion test, publication CSS and revision history.
Upload figures/model_abilities.svg with huggingface_hub
Upload figures/fit_scatter.svg with huggingface_hub
Upload figures/information_ranked.svg with huggingface_hub
Upload figures/item_parameters.svg with huggingface_hub
Upload index.html with huggingface_hub
Upload CYBENCH_IRT_REPORT.html with huggingface_hub
Upload figures/item_parameters.svg with huggingface_hub
Upload CYBENCH_IRT_REPORT.html with huggingface_hub
Upload index.html with huggingface_hub
Upload CYBENCH_IRT_REPORT.md with huggingface_hub
Upload index.html with huggingface_hub
Upload CYBENCH_IRT_REPORT.html with huggingface_hub
Upload figures/item_parameters.svg with huggingface_hub
Sync CyBench IRT pilot report details and figures
Sync CyBench IRT pilot report formatting fixes
Upload updated CyBench IRT pilot report with archive
initial commit