About eval_results.

[lixiangcog]

python data_process/train_dataset/order_easy_to_hard.py \
    --dataset_name UCSC-VLAA/MedVLThinker-m23k-tokenized \
    --split train \
    --results_jsonl_path outputs/estimate_pass_rate/Qwen2.5-VL-7B-Instruct/med-vlm-m23k/eval_results.jsonl \
    --save_to_disk_path data/local/med-vlm-m23k-easy_to_hard

Could you please explain how to generate the eval_results.jsonl file? I have reviewed the entire project but could not find the functionality for generating this file.

[lixiangcog]

I can generate this. 👆

[xk-huang]

Please follow the instructions in https://github.com/UCSC-VLAA/MedVLThinker/blob/main/docs/data.md#prepare-pass-rate-filter-by-pass-rate-and-order-to-easy-to-hard

The merged result is generated by https://github.com/UCSC-VLAA/MedVLThinker/blob/main/eval/merge_results.py