SOTA Long-context recall benchmarks and papers with code

Benchmarks

Dataset Name	SOTA Method	Metric
Long-context Recall-intensive Tasks Suite SQuAD, FDA, SWDE	Qwen3	SQuAD Recall Accuracy72.3	48	25d ago
NIAH Single-3	Qwen3	Recall @ 32K Context100	4	5mo ago
NIAH Single 2	Qwen3	Recall @ 32K Context1	4	5mo ago
NIAH Single-1	Qwen3	Recall @ 32K100	4	5mo ago

Showing 4 of 4 rows