modelscope / modelscope/evalscope

Question: standard vs CoT prompting in API calls

Open
#930 5 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

enhancement
Dominant language
Python
Stars
3.4k
Forks
489
Avg merge
1d 7h
Merged PRs (30d)
87

Description

Hi, is there currently a way to evaluate the same dataset using different prompt templates (e.g., MultipleChoiceTemplate.SINGLE_ANSWER and MultipleChoiceTemplate.SINGLE_ANSWER_COT) via the API or CLI call, without redefining the benchmark adapter?

This would make it much easier to compare LLMs’ performance on standard vs. CoT prompting styles using the same dataset and evaluation logic.

Warm regards,
Yang

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by tracing the API and CLI entry points for benchmark evaluation and how benchmark adapters select prompt templates. Check whether MultipleChoiceTemplate.SINGLE_ANSWER and SINGLE_ANSWER_COT can be selected without redefining an adapter. Done means the same dataset and evaluation logic can run with both templates through a documented API or CLI option.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
api, cli, machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.