List your company

Multi-document Question Answering on MDQA (Accuracy Thresholds)

72.6Accuracy @ 0% ThresholdAttention Buckets
67.19268.5967071.404Jun 26, 2026

Evaluation results

MethodLinks
Attention BucketsModel=Vicuna-7B-v1.52026.0672.661.460.660.859.663
Ms-PoEModel=Vicuna-7B-v1.52026.0672.661.461.8625963.5
MoICEModel=Vicuna-7B-v1.52026.0671.661.260.660.862.463.3
LPESModel=Vicuna-7B-v1.52026.0671.462.2626161.663.6
Positional InterpolationModel=Vicuna-7B-v1.52026.0671.259.658.856.456.260.4
BaselineModel=Vicuna-7B-v1.52026.0670.45855.455.460.459.9
Positional InterpolationModel=StableBeluga-7B2026.0669.658.658.26065.462.4
Attention BucketsModel=Qwen2.5-7B2026.0669.662.26360.26263.4
LPESModel=Qwen2.5-7B2026.0669.664.869.26365.466.4
BaselineModel=Qwen2.5-7B2026.0669.46162.658.663.663
Ms-PoEModel=Qwen2.5-7B2026.0669.461.863.460.261.463.2
Attention BucketsModel=StableBeluga-7B2026.0669.25959.859.267.463
LPESModel=StableBeluga-7B2026.0668.86060.86168.264.5
Positional InterpolationModel=Qwen2.5-7B2026.0668.66262.258.46463
Ms-PoEModel=StableBeluga-7B2026.0668.45760.26168.463
MoICEModel=Qwen2.5-7B2026.0668.461.2636163.863.5
BaselineModel=StableBeluga-7B2026.0667.859.259.659.468.262.8
MoICEModel=StableBeluga-7B2026.0667.46060.26068.663.2