List your company

ResearchFormerly WizWand

ELI5

Benchmarks

TaskDatasetSOTA resultTrendResults

Long Form Question Answering

ELI5

31.15ROUGE-L

57

Long Form Question Answering

ELI5 (test)

27.13ROUGE-L

54

eli5-category (test)

1.313PPL

28

Dialectal alignment estimation

ELI5 informal

0.77Percentage of AmE Generations

20

Attributed Text Generation

ELI5

25.9Claim Correctness Score

19

ELI5

91.79Accuracy

16

Text generation quality and watermark detectability

ELI5

100AUC

16

ELI5

55.8Precision

15

Word-level length-controlled question answering

ELI5 (test)

14.1MAE

14

Question Answering

ELI5 Wiki-answerable

26.6ROUGE-L Score

14

Machine-Generated Text Detection

ELI5 Domain Generalization (DG-MGT)

91.79Accuracy

12

Sentence-level length-controlled question answering

ELI5 (test)

0.08MAE

12

Token-level length-controlled question answering

ELI5 (test)

10.33MAE

12

Long-Form Question Answering

ELI5 (val)

31.5F1

11

Sentence-level attribution

ELI5 (test)

81.9Citation Recall

10

Grounded Generation

ELI5 (test)

47.43Fluency (mauve)

10

Knowledge-intensive generation

ELI5 (dev)

26.6ROUGE-L Score

9

Distortion-Free Watermark Evaluation

ELI5 16-bit bitstrings LLAMA3.1-8B

77.8Message Accuracy

8

Long-form Question Answering

ELI5 KILT (test)

25.4F1

8

Long-form Question Answering with Citations

ELI5

0.186Correctness

8

ELI5 KILT (test)

11Retrieval Precision

8

ELI5

20.75R-L

8

Paraphrasing Robustness

ELI5 prompts 32-bit payload Llama3.1-8B (test)

50.9Bit Accuracy

7

Synonym Substitution Robustness

ELI5 prompts 32-bit payload Llama3.1-8B (test)

70.7Bit Accuracy (10% Substitution)

7

Word Deletion Robustness

ELI5 prompts 32-bit payload Llama3.1-8B (test)

70.7Bit Accuracy (10% Deletion)

7