List your company

Multi-modal Task Evaluation on MM-MergeBench Unseen Tasks

79.3AVQA ScoreRobustMerge
50.512857.986465.4672.9336Mar 6, 2026

Evaluation results

MethodLinks
RobustMergeSetting=Unseen/Zero-shot generalization2026.0379.345.799.2317.6237.99
DC-MergeSetting=Unseen/Zero-shot generalization2026.0379.350.2410.8818.9439.84
Iso-CTSSetting=Unseen/Zero-shot generalization2026.0378.4844.8111.2113.9937.12
TSV-MSetting=Unseen/Zero-shot generalization2026.0377.2942.7211.1213.836.23
Multi TaskSetting=Unseen/Zero-shot generalization2026.0376.3341.398.3418.236.06
Task ArithmeticSetting=Unseen/Zero-shot generalization2026.0374.7837.377.5213.5733.31
PCB-MergingSetting=Unseen/Zero-shot generalization2026.0374.5736.287.8415.4433.53
DARE-MergingSetting=Unseen/Zero-shot generalization2026.0373.7537.677.5613.6233.15
TIES-MergingSetting=Unseen/Zero-shot generalization2026.0373.4338.447.4713.2333.14
WUDI-MergingSetting=Unseen/Zero-shot generalization2026.0368.5641.5711.1611.833.27
ZeroshotSetting=Unseen/Zero-shot generalization2026.0351.6228.275.9815.0125.22