feat: new benchmark OlympiadBench #16
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Kass123777
reviewed
Apr 1, 2025
Kass123777
reviewed
Apr 1, 2025
Lavezlyn
reviewed
Apr 4, 2025
| data_files: null | ||
| <<: *task_defaults | ||
| answer_extractor: | ||
| - name: match_multi-choice_and_open-ended |
Contributor
There was a problem hiding this comment.
The name should match the function.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
Add new benchmark OlympiadBench
The OlympiadBench was published at February 2024. Similar to Mathvision, they only tested a few closed-source models.
Their code does not include prompts or hyperparameter settings for existing open-source models.
Regarding prompts, we use the prompt for Qwen-VL from their OlympiadBench GitHub.
Current test results are as follows:
For Qwen/Qwen2-VL-72B-Instruct, Qwen's official score is 11.2
Using greedy decoding, I achieved a score of 0.11104347826086956.
The scores for qvq are still under testing, but qvq as a reasoning model requires very long inference time. Additionally, among open-source models, I could only find scores for Qwen2VL-72B. Therefore, after testing Qwen2VL-72B and obtaining scores close to the Qwen official report, I believe the current results are acceptable.
Furthermore, OlympiadBench has two types of problems: Open-ended questions with ground truth, and Theorem proof problems. Although this dataset provides Theorem proof problems, they mentioned in their paper that they cannot automatically evaluate Theorem proof problems, so currently only Open-ended questions have been integrated.
Motivation and Context
Why is this change required? What problem does it solve?
If it fixes an open issue, please link to the issue here.
You can use the syntax close #1314520 if this solves the issue #15213
Types of changes
What types of changes does your code introduce? Put an x in all the boxes that apply:
Checklist
Go over all the following points, and put an x in all the boxes that apply.
If you are unsure about any of these, don't hesitate to ask. We are here to help!