stanford-crfm / helm

Holistic Evaluation of Language Models (HELM), a framework to increase the transparency of language models (https://arxiv.org/abs/2211.09110). This framework is also used to evaluate text-to-image models in Holistic Evaluation of Text-to-Image Models (HEIM) (https://arxiv.org/abs/2311.04287).
https://crfm.stanford.edu/helm
Apache License 2.0
1.89k stars 244 forks source link

Fix GPT4V Evaluator Out of Option Range Issue #2677

Closed ImKeTT closed 4 months ago

ImKeTT commented 4 months ago

This PR is to fix the GPT-4V auto-evaluator out of option range issue on the Mementos scenario.