[P1] Questions about evaluate.py and Concept500-HO dataset

[clarenceluo78]

Thank you for the interesting project and code! I have a couple of questions:

1. When running demo.sh, the script axbench/scripts/evaluate.py seems incomplete:

   • eval_latent does not return any results.
   • A small bug occurs with KeyError: 'LsReFT_perplexity' at line 556 in eval_steering:

     for concept_id, evaluator_str, model_str, result, lm_report, lm_cache, current_df in executor.map(...)

2. Will the Concept500-HO dataset from the HyperSteer paper be released?
   Currently, I only see the Concept500 dataset in the repository.

Thanks again for sharing this work!