feat: add Multi-Level Existence Benchmark (MLE-Bench)#1228
Merged
Luodian merged 2 commits intoEvolvingLMMs-Lab:mainfrom Mar 7, 2026
Merged
feat: add Multi-Level Existence Benchmark (MLE-Bench)#1228Luodian merged 2 commits intoEvolvingLMMs-Lab:mainfrom
Luodian merged 2 commits intoEvolvingLMMs-Lab:mainfrom
Conversation
Adds evaluation support for the Multi-Level Existence Benchmark (MLE-Bench), introduced in the ICLR 2026 Oral paper: Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-training Junlin Han, Shengbang Tong, David Fan, Yufan Ren, Koustuv Sinha, Philip Torr, Filippos Kokkinos ICLR 2026 (Oral) https://openreview.net/forum?id=pfw176o1YJ Project page: https://junlinhan.github.io/projects/lsbs/ MLE-Bench evaluates fine-grained visual perception in multimodal models using 4-choice questions about object existence, categorised by the target object's relative size (proportion of image pixels occupied): - small (existence_0-30): 732 samples, objects occupying 0-30% of image - medium (existence_30-60): 698 samples, objects occupying 30-60% of image - large (existence_60-100): 431 samples, objects occupying 60-100% of image Dataset: https://huggingface.co/datasets/JunlinHan/Multi-Level_Existence_Bench Tasks added: - mle_bench : full evaluation (1,861 samples) - mle_bench_small : small-object subset - mle_bench_medium : medium-object subset - mle_bench_large : large-object subset Metrics: per-category accuracy (small / medium / large) + macro-average
Contributor
|
@claude please lint this |
|
Claude Code is working… I'll analyze this and get back to you. |
Luodian
approved these changes
Mar 7, 2026
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.
Adds evaluation support for the Multi-Level Existence Benchmark (MLE-Bench), introduced in the ICLR 2026 Oral paper:
Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-training
Junlin Han, Shengbang Tong, David Fan, Yufan Ren, Koustuv Sinha, Philip Torr, Filippos Kokkinos
ICLR 2026 (Oral) — https://openreview.net/forum?id=pfw176o1YJ
Project page: https://junlinhan.github.io/projects/lsbs/
Dataset: https://huggingface.co/datasets/JunlinHan/Multi-Level_Existence_Bench
Summary
In scope
lmms_eval/tasks/mle_bench/withutils.pyand YAML configsmle_bench(full),mle_bench_small,mle_bench_medium,mle_bench_largeOut of scope
Validation
python -m lmms_eval --model openai --model_args model=glm-4v-flash,base_url=https://open.bigmodel.cn/api/paas/v4/ --tasks mle_bench --limit 100| sample size:N=100(stratified) | key metrics: small=91.2%, medium=93.9%, large=93.9%, overall=93.0% | result: passRisk / Compatibility
Type of Change