Creating dataframe with rank and "correct_term" sometimes has (silent) unexpected behavior

[leokim-l]

For example in:
https://github.com/monarch-initiative/pheval.llm/blob/main/out_openAI_models/multimodel/gpt-4o/full_df_results.tsv

Line 1498 (in total in 110 total lines/rows) the correct_term column contains the rank instead of the OMIM. Phenopackets seem fine (although older versions may have had bugs). I suspect this issue to be in some measure in all output we have.

Probably some issue here or somewhere farther below:

pheval.llm/src/malco/post_process/ranking_utils.py

Line 81 in 5265eb3

label_to_correct_term = answers.set_index("label")["term"].to_dict()