Data files

[xiye17]

Thanks for the great work and for sharing the dataset. I am trying running the inference.

It seems in scripts/eval/inference.py ln 160 the script is reading a csv input, which I cannot find in the repo or huggingface. Can you provide the test file?

CLIPPER/scripts/eval/inference.py

Line 160 in 4cb8019

data_input = f"../../data/benchmark/{args.data_input}.csv"

Also I'd like to ask a question on the results. In Figure 1 of the paper, Qwen2.5-Instruct-7B is achieving a very good 51% on the dataset, is it single claim accuracy or paired-claim accuracy?

Many thanks!