open-compass / open-compass/VLMEvalKit

difference sample number with HallusionBench's paper

Open
#1,037 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Awaiting Confirm
Dominant language
Python
Stars
4.4k
Forks
768
Avg merge
2d 27m
Merged PRs (30d)
18

Description

I try to evaluate models on HallusionBench.
The number of hallusionbench in lmevalkit is 951, which is 1129 in HallusionBench's huggingface.
I wonder why they have different number.
What's more, I get only about 30 score of seed1.5vl on hallusionbench, which is far from 60.3 in the seed1.5vl paper.
Could someone get the similar result with me?
Thanks.

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 by comparing the HallusionBench sample count in VLMEvalKit with the 1129 samples on HallusionBench's Hugging Face version. Then reproduce the Seed1.5VL evaluation and compare the roughly 30 score with the 60.3 reported in its paper. Done means the dataset and score discrepancies are explained or reproducibly confirmed.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning, testing-qa
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.