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

MRCR

Benchmarks

Task NameDataset NameSOTA ResultTrend
Long ContextMRCR
Score89.7
25
Long Context UnderstandingMRCR
Accuracy75.3
15
Long-context reasoningMRCR OOD 8-128K
Accuracy (8-16K)96.6
10
Long-context reasoningMRCR ID 4-8K
Accuracy99.8
10
Multi-needle RetrievalMRCR
MRCR 2-Needle Success38
9
Long-context & Multi-turn DialogueMRCR 16K-256K
Score80.37
8
Multi-round coreference resolutionMRCR
2-Needle Score81.6
8
Long Context RetrievalMRCR v2
Accuracy66.4
6
Complex retrieval and positional sortingMRCR 512K~1M
Score46.88
6
Complex retrieval and positional sortingMRCR 128K~512K
Score53.98
6
Multi-Round Chat RetrievalMRCR
Accuracy92.3
6
Long Context PerformanceMRCR v2
Score89.4
5
ReasoningMRCR v2
Score94.8
5
Multi-round Code RepairMRCR
pass@10.77
5
Long Context PerformanceMRCR 8-needle v2
Score79
4
Multi-needle retrievalMRCR-8needle 100-example
Sequence Match Accuracy27.9
4
Showing 16 of 16 rows