Xsum
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
XSum (test)
24.86ROUGE-1
5
XSum subset of 50 documents (test)
0.431Score
5
XSum
61.72FactKB Score
4
XSum 100 samples
37SDE
4
XSum
53.09CV Topic Coherence Score
4
XSum
76.7AUC
4
XSum
43.1ROUGE-2
4
XSum (test)
0.232ROUGE-2
4
XSUM 40K
13BLEU
4
XSum
57.5QA Score
4
XSum
34.08BERTScore F1
3
XSum Llama2-13B
80.67TPR@FPR=1%
3
XSum GPT-5-Chat
31.33TPR @ FPR=1%
3
XSum
0.742AUC (LOSS)
3
XSUM Llama dpo-1-ling (test)
32.4TPR @ 0.01 FPR
3
XSUM Llama dpo-1 (test)
0.62TPR @ 0.01 FPR
3
XSUM Llama (test)
93.2TPR @ 1% FPR
3
XSum (val)
1,736.88Peak VRAM (MB)
2
XSum (val)
17.9ROUGE-2
2
XSum
7.15ROUGE-L
2
XSum 100 sample subset (test)
6Informativeness (Pairwise Lose)
2
XSum
48.3Win Rate
1
GEM XSum en (test)
—ROUGE-2
0