Writing
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Writing Generated by Claude3 (test)
99.5AUROC
15
Writing Generated by GPT-4 (test)
0.977AUROC
15
Writing Generated by ChatGPT (test)
0.992AUROC
15
Writing 10-shot
91.3Accuracy
10
Writing 5-shot
87Accuracy
10
Writing 3-shot
75.1Accuracy
10
Writing N=4k
94.2R@1
7
Writing 5-shot
7.71Training Time (mins)
5
Writing 10-shot
92.22GPU Utilization (%)
5
Writing 5-shot
85.47GPU Utilization
5
Writing 3-shot
75.14GPU Utilization
5
Writing
467.5Watch Latency (ms)
4
Writing Synthetic (test)
87.6AUROC
3
Writing
1,000Ideas Accepted
3
Writing Unconstrained
22.1F1 Score
3
Writing Category-controlled top-K
14.2F1 Score
3