MiMo-VL-7B-SFT-2508: benchmark results
Provider: Xiaomi. Access: Open.
Aggregate Rating 1576 ± 28, rank #1315 of 10916 rated models, from 30 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HyperTrack | 77.98 | Step Exact Match (%; predicted action type and parameters bo | 100 |
| RobotEQ-Video - Real Videos | 76.4 | Real-subset Avg (%; action-properness judgment: given an ego | 83 |
| M-JudgeBench - Short CoT Cross-Model Pairs | 68.69 | pairwise accuracy (%) on the 297 pairs where a correct and a | 63.2 |
| RobotEQ-Video | 73.8 | A-Avg: mean of the synthetic-subset and real-subset averages | 61.7 |
| RobotEQ-Video - Synthetic Videos | 73.3 | Synthetic-subset Avg (%; action-properness judgment: given a | 59.6 |
| EPIC-Bench - Visual Matching | 36.89 | Navigation score (0-100) on visual matching (localizing an o | 56.8 |
| EPIC-Bench - Embodied Compositional Attributes | 32.98 | Target localization score (0-100) on the embodied compositio | 52.3 |
| EPIC-Bench - Placement Region | 32.82 | Manipulation score (0-100) on placement region (where an obj | 45.5 |
| M-JudgeBench - Incidental Mistake Detection | 73.81 | pairwise accuracy (%) on the 252 pairs where the rejected ch | 44.7 |
| M-JudgeBench - Logical Fallacy Detection | 88.65 | pairwise accuracy (%) on the 238 pairs where the rejected ch | 42.1 |
| EPIC-Bench - Affordance Region | 4.91 | Manipulation score (0-100) on affordance region localization | 40.9 |
| EPIC-Bench - Feasible Path | 29.14 | Navigation path score (0-100) on feasible path planning (a p | 34.1 |
Interactive version: theaggregate.ai/model?slug=mimo-vl-7b-sft-2508 · How It Works · Data refreshed daily, snapshot 2026-11-08.