search

EQ-bench / EQ-Bench

A benchmark for emotional intelligence in large language models
MIT License
180 stars 13 forks source link

Error in calculating revise answer score #25

Closed impact-rm closed 3 months ago

impact-rm commented 4 months ago

When I call the eq-bench with -revise enabled it returns "! Error: 4 emotions were not returned" and revised_answers: {}

impact-rm commented 4 months ago

The problem is solved by increasing the COMPLETION_TOKENS (which is set to 60) in eq_bench_utils.py. As it's the value parsed to max_tokens of openai.

sam-paech commented 3 months ago

Ah, thank you. I overlooked this, will fix.