List your company

Question Answering on 8 QA Benchmarks Suite (OBQA, PIQA, ARC-E, ARC-C, HellaSwag, SciQ, WebQs, RACE) (test)

38.8OBQACRaSh
18.83224.01629.234.384Jul 6, 2025

Evaluation results

MethodLinks
CRaShBackbone=OPT-6.7B, Compression Category=Training-free Compression, Setting=Plug-in2025.0738.87870.736.353.495.326.145.255.3
ScaleOTBackbone=LLAMA-7B, Compression Category=Post-training Compression, Setting=Plug-in2025.0737.479.773.242.358.195.733.745.658.2
GradOTBackbone=OPT-6.7B, Compression Category=Training-free Compression, Setting=Plug-in2025.0734.8787034.852.495.332.745.155.4
CRaShBackbone=LLAMA-7B, Compression Category=Training-free Compression, Setting=Plug-in2025.0734.68071.341.858.495.129.845.657.1
OTBackbone=OPT-6.7B, Compression Category=Training-free Compression, Setting=Plug-in2025.0733.877.766.833.952.191.923.944.153
GradOTBackbone=LLAMA-7B, Compression Category=Training-free Compression, Setting=Plug-in2025.0733.880.572.841.757.995.438.344.858.2
ScaleOTBackbone=OPT-6.7B, Compression Category=Post-training Compression, Setting=Plug-in2025.0733.278.170.135.352.295.733.945.355.5
OTBackbone=LLAMA-7B, Compression Category=Training-free Compression, Setting=Plug-in2025.073378.869.63957.483.527.34454.1
EmulatorBackbone=LLAMA-7B, Evaluation Protocol=Fine-tuning (FT)2025.0730.676.363.332.649.491.829.643.252.1
EmulatorBackbone=OPT-6.7B, Evaluation Protocol=Fine-tuning (FT)2025.0728.473.664.530.94693.123.441.550.2
Full ModelBackbone=LLAMA-7B, Evaluation Protocol=Zero-shot (ZS)2025.0728.278.367.338.256.489.704049.8
Full ModelBackbone=OPT-6.7B, Evaluation Protocol=Zero-shot (ZS)2025.0727.676.265.530.650.590.18.838.248.4
EmulatorBackbone=LLAMA-7B, Evaluation Protocol=Zero-shot (ZS)2025.072372.152.83045.272.6032.341
EmulatorBackbone=OPT-6.7B, Evaluation Protocol=Zero-shot (ZS)2025.0719.668.853.624.738.1832.431.640.2