Average
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Average Cities D, E, F (Cold-start)
656.8Score
3
Average WGM, OQA, WVS
86.2Opinion Alignment
3
Average
3,270Pixel Matching
3
Average Recon, Tartan Drive, Scand, Sacson (val)
2.975ATE
3
Average 25 Tasks
85.93Average Score
3
Average Videos 7-9
99.99ASR
3
Average
30.698PSNR
3
Average
0.78Word-level ROC AUC
3
Average
78Word-level ROC AUC
3
Average All Programs
6,684Peak Memory (MB)
2
average of 4 datasets
92.9Delta Accuracy 1
2
Average ID
45.66Avg@64
2
Average
2.172RMSE
2
Average 8 datasets
9.5WER
2
Average (MMLU, GSM8K, MBPP)
78.84Accuracy
2
Average HMMT2025, GPQA-D, AIME2024, AIME2025, BRUMO2025
77.84Average Accuracy
2
Average
24.5NDCG@25
2
Average 8 image & 4 video datasets
75.7Top-1 Accuracy
2
Average OOD
19.39Avg@64
1
Average
13.6MSE
1
Average (6 datasets) macro-average (test)
57.68EM
1
Average Internal + NH 3.0 (All Datasets)
—BLEU
0
Average over 5 datasets v1 (test)
—Base Class Accuracy
0