List your company

Multiple-choice Question Answering on MMLU-ProX

82.6Average AccuracyORACLE
21.55237.40153.2569.099Apr 18, 2026Apr 19, 2026Apr 20, 2026Apr 21, 2026Apr 22, 2026

Evaluation results

MethodLinks
ORACLEStrategy=Upper bound selection, Backbone=DeepSeek-v3.12026.0482.685.485.58584.283.385.582.880.670.8——
CLASSIFIER (Ours)Strategy=Learned Routing (XGBoost), Backbone=DeepSeek-v3.12026.0477.980.881.380.579.278.780.779.57664.6——
TRANSLATEStrategy=Translate-then-inference, Backbone=DeepSeek-v3.12026.0477.780.38180.37978.580.579.575.664.5——
NATIVEStrategy=Native inference, Backbone=DeepSeek-v3.12026.0469.280.580.879.777.975.480.135.769.343.4——
ORACLEBase Model=Llama3.3-70B2026.0465.471.372.371.467.364.47263.962.743.6——
XGBOOST (Ours)Base Model=Llama3.3-70B2026.0458.56467.166.16155.766.655.954.236——
MLP (Ours)Base Model=Llama3.3-70B2026.0458.364.36766.26156.566.651.356.135.8——
TRANSLATEBase Model=Llama3.3-70B2026.0453.957.859.958.754.552.959.65649.936.2——
NATIVEBase Model=Llama3.3-70B2026.0453.663.86665.457.853.465.540.251.618.8——
Naive Fine-tuningBackbone=LLaMA-3.1-Instruct-8B, Number of random seeds=32026.0425.3———————————
COMPASS-ECDABackbone=LLaMA-3.1-Instruct-8B, Number of random seeds=32026.0425.2———————————
Full RetrainingBackbone=LLaMA-3.1-Instruct-8B, Number of random seeds=32026.0424.6———————————
EWCBackbone=LLaMA-3.1-Instruct-8B, Number of random seeds=32026.0424.5———————————
Random RehearsalBackbone=LLaMA-3.1-Instruct-8B, Number of random seeds=32026.0423.9———————————