MMDM - Multi Select Accuracy: leaderboard
MMDM - Multi Select Accuracy: Hard accuracy (%). MMDM arXiv 2611.14309v1: author-run decision-head or single-decoding-step models, no generated reasoning/adaptive thinking. Hard accuracy counts failed requests and boundary ties wrong. Native option caps/context differ by model and are pinned in Table A5. Soft Brier (0-2) and score normalized MAE (0-1) are valid-native-response errors, not full-cohort accuracy. Model-specific SB-F from Table 2 is withheld in favor of Table A10 fixed common 284-native-question SB-F (0-1). Source-native released artifact and service version identities remain distinct.
Metric: Hard accuracy (%). Source: arxiv.org. Saturation forecast: Around August 2028. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | JEV-27B-VL | 20.22 | #219 |
| 2 | GPT-6 Luna | 12.01 | #93 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mmdm-multi-select-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-12.