Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

MemBench

Benchmarks

Task NameDataset NameSOTA ResultTrend
Memorability Feedback GenerationMemBench
IR0.85
13
Question AnsweringMemBench High
Accuracy77.32
12
Question AnsweringMemBench Low
Accuracy88.78
12
Reflective memoryMemBench 100k context
Accuracy86.3
10
Long-context memory evaluationMemBench
Recall87.5
10
Reflective memoryMemBench 10k context
Reflective Accuracy84.3
8
Text-to-Image GenerationMemBench 3000 Memorized Prompts 1.0 (test)
SSCD (Target)50.3
5
Text-to-Image GenerationMembench 3,000 prompts 1.0
SSCD0.79
3
Feedback Quality EvaluationMemBench human study
Clearness4.19
2
Showing 9 of 9 rows