question regarding judgement

[MarSan2001]

the response part of the debaters both incorrect and correct gives "Round ended so no comment made. Judge - do not count this round" why could this be

[jplhughes]

Hi @MarSan2001 - this will be because of a context length limit issue with the LM you are using. If it hits the cap of the model then it discards the round it got to. What LM are you using?

[MarSan2001]

gpt-4-1106-preview is the LM I am using. Is it possible to increase the cap

[jplhughes]

hmm, it should never trigger for that

llm_debate/core/agents/debater_quality.py

Line 260 in f9c71d1

if (

only ft models or 3.5-turbo