ulab-uiuc / research-town

A platform for AI researchers to study automatic AI research
http://docs.auto-research.dev
Apache License 2.0
12 stars 1 forks source link

[FEAT]: polish eval and agent prompt #385

Closed Monstertail closed 1 month ago

Monstertail commented 1 month ago

Description

polish eval prompter for quality evaluators.

Additional Information

No response

Monstertail commented 1 month ago

@lwaekfjlk shall we separate the prompt to get a score in 1 dimension (e.g. 7 for the first criterion) other than all dimension scores (e.g. [7,7,8] for all three criteria )in each evaluation prompting?