weitaikang/

Weitaikang_bench_paper

The evidence behind the paper: every judge call, every human annotation, every ablation reading, every per-unit score. One of three repositories:

Repository Contents

archive you are here judge calls, human study, ablations, per-unit scores, figures

benchmark Weitaikang_bench_data 2.8 GB — chains, images, rubrics, frame-track index

code Weitaikang_bench_github the evaluation package and all eleven baselines

The… See the full description on the dataset page: https://huggingface.co/datasets/weitaikang/Weitaikang_bench_paper.