feat(benchmarks): add hmmt_nov25 benchmark - #1636
Merged
Merged
Conversation
Add HMMT November 2025 as a first-class math benchmark, following the existing hmmt25/hmmt26 adapters.
Yunnglin
approved these changes
Aug 26, 2026
Yunnglin
left a comment
Yunnglin
left a comment
Collaborator
There was a problem hiding this comment.
LGTM! ModelScope loading, documentation, mock/real API E2E, and scoring have been verified.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Add one new math benchmark, following the existing HMMT family (
hmmt25/hmmt26):hmmt_nov25, dataset idevalscope/hmmt_nov_2025, 30 problems) — Harvard-MIT Mathematics Tournament November 2025. This is a different contest from HMMT February 2025 (hmmt25).Source: MathArena/hmmt_nov_2025 (CC BY-NC-SA 4.0). Adapter
dataset_iduses the ModelScope convention, same asevalscope/hmmt_feb_2025andevalscope/hmmt_feb_2026.Changes
hmmt_nov25adapter underevalscope/benchmarks/hmmt/, reusingextract_hmmt_answertest_hmmt_nov25next to the existing HMMT tests_metaJSONREADME.md/README_zh.mdTest plan
hmmt_nov25and load 30 local samplesextract_answeron a boxed gold answerevalscope eval --eval-type mock_llm --datasets hmmt_nov25 --limit 1withlocal_pathdeepseek-chatvia official API: Accuracy 50% (1/2)test_hmmt_nov25against ModelScopeevalscope/hmmt_nov_2025(dataset not published yet)evalscope eval --datasets hmmt_nov25 --limit 1after the ModelScope dataset is uploadedNotes
evalscope/hmmt_nov_2025is not on ModelScope yet. Until it is uploaded, run withlocal_pathor--dataset-hub huggingface+dataset_id=MathArena/hmmt_nov_2025.