List your company

Long-context evaluation on RULER 32k (matched-train holdout)

48.7AccuracyStill
17.525.633.741.8Jun 5, 2026

Evaluation results

MethodLinks
StillK (Cache budget)=2048, Number of seeds=32026.0648.7
StillK (Cache budget)=1024, Number of seeds=32026.0645.2
StillK (Cache budget)=328, Number of seeds=32026.0639.2
StillK (Cache budget)=512, Number of seeds=32026.0638.4
StillK (Cache budget)=164, Number of seeds=32026.0634.5
StillK (Cache budget)=82, Number of seeds=32026.0629.1
KV-DistillK (Cache budget)=2048, Number of seeds=32026.0626.5
KV-DistillK (Cache budget)=1024, Number of seeds=32026.0624
KV-DistillK (Cache budget)=512, Number of seeds=32026.0622.1
KV-DistillK (Cache budget)=328, Number of seeds=32026.0621.2
KV-DistillK (Cache budget)=164, Number of seeds=32026.0619
KV-DistillK (Cache budget)=82, Number of seeds=32026.0618.7