AVG
Benchmarks
TaskDatasetSOTA resultTrendResults
AVG.
47.5EM
48
Avg.
6.73Speedup
32
AVG GSM8K MATH-500 AIME 2024 OlympiadBench GPQA Diamond
78.4Accuracy
24
Avg. (held out)
62.8Accuracy
24
Avg across SYSU, LEVIR, GVLM, CLCD, OSCD
84.8Precision
23
AVG Reasoning Suite
77.4Accuracy
18
AVG. DICM, MEF, LIME, NPE, VV
3.589NIQE
17
AVG
54.373NDCG@5
14
AVG
351TPF
14
AVG. Aggregate of NQ, TQA, HQA, 2WIKI (test)
42.5EM
14
AVG AMC 2023, AIME 2025, HumanEval, LiveCodeBench
83.4Accuracy
12
AVG IEMOCAP, MELD, EmoryNLP
60.91W-F1
11
Avg (all)
0.215AURC (10^-2 Scale)
11
Avg 1K
0.248AURC (Scale 10^-2)
11
AVG (test)
8.3City Acc (25km)
10
AVG
0.897AUC
10
Avg. across four cases Combined (test)
40.63PSNR
9
AVG Across All Benchmarks
12.89Throughput
8
AVG 12 benchmarks
91.33Word Accuracy
8
Avg-14
76.19Top-1 Clean Accuracy
7
Avg. All five datasets
25.5Median Error (Med)
5
Avg.
70.6GRA
4
Avg (News, Flores, Subtitle, Travel) German-English (test aggregate)
87.77DA
4