lm-sys / lm-sys/FastChat

Wrong reference answer in MT-Bench Question 104

Open
#3,053 1 comment 2 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
39.5k
Forks
4.8k
PR merge metrics
No merged PRs in 30d

Description

In MT-Bench, both answers to question_id=104 are wrong

https://github.com/lm-sys/FastChat/blob/ed6735d84a198325e1f6a155976987bc75e1f14a/fastchat/llm_judge/data/mt_bench/question.jsonl#L24

https://github.com/lm-sys/FastChat/blob/81b80fd22708204f14fa7a5177cfa456f0660ebc/fastchat/llm_judge/data/mt_bench/reference_answer/gpt-4.jsonl#L4-L4

Both turns provided by GPT-4 are wrong, because the first turn should be that David has no brother and the other turn should be one brother.

Contributor guide

No contributing guide indexed for this repository

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 with fastchat/llm_judge/data/mt_bench/question.jsonl at line 24 and fastchat/llm_judge/data/mt_bench/reference_answer/gpt-4.jsonl at line 4. Compare both turns for question_id=104 with the issue’s expected answers, update the incorrect reference data, and verify that the entry now matches the question.

Written by the indexing model from the issue text.

Assessment

Domain
machine-learning, testing-qa
Issue type
Bug
Difficulty
1/5
Estimated time
1-3 hours
Activity status
Stale
Clarity
Clearly specified
Newbie friendliness
58/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.