Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Needle-in-a-Haystack (NIAH) Retrieval on RULER 4K
Loading...
100
MV Accuracy
MemDLM (Train & Inference)
87.416
90.683
93.95
97.217
Mar 23, 2026
MV Accuracy
VT Accuracy
CWE Accuracy
Updated 25d ago
Evaluation Results
Method
Method
Links
MV Accuracy
VT Accuracy
CWE Accuracy
MemDLM (Train & Inference)
Backbone=LLaDA2.1
2026.03
100
90.16
80.96
MemDLM (Train-Only)
Backbone=LLaDA2.1
2026.03
99.65
88.56
78.96
Standard MDLM
Backbone=LLaDA2.1
2026.03
98.65
59.32
72.34
MemDLM (Train & Inference)
Backbone=LLaDA-MoE
2026.03
96.8
99.6
83.7
MemDLM (Train-Only)
Backbone=LLaDA-MoE
2026.03
95.85
99.52
75.72
Standard MDLM
Backbone=LLaDA-MoE
2026.03
87.9
99.4
69.1
Feedback
Search any
task
Search any
task