List your company

Long-context language modeling on InfiniteBench (test)

34.82En QA ScoreLlama 3.1 8B Instruct
-0.93528.347417.6326.9126Nov 5, 2025Nov 16, 2025Nov 27, 2025Dec 9, 2025Dec 20, 2025Dec 31, 2025Jan 12, 2026

Evaluation results

MethodLinks
Llama 3.1 8B InstructBase Model=Llama 3.1 8B Instruct, KV Cache Management Method=Full Attention, KV Cache Length/Budget=Full2025.1134.82——————10099.3287—
SAGE-KVBase Model=Llama 3.1 8B Instruct, KV Cache Management Method=SAGE-KV, KV Cache Length/Budget=8k2025.1134.69——————10096.11—
SnapStreamBase Model=Llama 3.1 8B Instruct, KV Cache Management Method=SnapStream, KV Cache Length/Budget=8k2025.1132.44——————10092.259—
StreamingLLMBase Model=Llama 3.1 8B Instruct, KV Cache Management Method=StreamingLLM, KV Cache Length/Budget=8k2025.1127.57——————6.786.442.4—
SnapKVModel=Llama-3.1-8B-UL, KV Budget=20482026.011.0110.990.990.991.011.011.021.011.011
SnapKVModel=Qwen2.5-7B, KV Budget=20482026.0111111111111
ASL_2passModel=Qwen2.5-7B, KV Budget=20482026.010.830.80.90.850.840.830.720.760.70.730.82
ASLModel=Qwen2.5-7B, KV Budget=20482026.010.820.80.90.850.850.820.730.770.70.730.82
ASL_2passModel=Llama-3.1-8B-UL, KV Budget=20482026.010.740.580.770.710.710.690.670.480.620.680.69
ASLModel=Llama-3.1-8B-UL, KV Budget=20482026.010.730.580.760.710.710.670.670.470.60.680.68
FastKVModel=Qwen2.5-7B, KV Budget=20482026.010.530.540.540.530.540.540.540.540.530.530.53
GemFilterModel=Qwen2.5-7B, KV Budget=20482026.010.530.540.530.530.530.530.530.540.520.530.53
FastKVModel=Llama-3.1-8B-UL, KV Budget=20482026.010.50.50.510.50.510.50.50.520.50.50.5
GemFilterModel=Llama-3.1-8B-UL, KV Budget=20482026.010.440.440.440.430.440.450.440.450.440.440.44