List your company

Long-context retrieval on Needle-in-a-Haystack (test)

100AccuracyFullKV
37.974454.077270.1886.2828Mar 12, 2026Mar 15, 2026Mar 18, 2026Mar 22, 2026Mar 25, 2026Mar 28, 2026Apr 1, 2026

Evaluation results

MethodLinks
FullKVLLM=Llama3-8B-Instruct, KV Cache Size=Full2026.03100
YOCO2026.04100
YOCO-U2026.04100
DapQLLM=Llama3-8B-Instruct, KV Cache Size=2562026.0399.46
RINS2026.0499
FullKVLLM=Llama3.1-8B-Instruct, KV Cache Size=Full2026.0398.02
FullKVLLM=Qwen3-8B, KV Cache Size=Full2026.0396.52
DapQLLM=Llama3-8B-Instruct, KV Cache Size=1282026.0395.75
FullKVLLM=Qwen2.5-7B-Instruct, KV Cache Size=Full2026.0394.23
PyramidKVLLM=Llama3-8B-Instruct, KV Cache Size=2562026.0393.94
DapQLLM=Qwen3-8B, KV Cache Size=2562026.0391.73
SnapKVLLM=Llama3-8B-Instruct, KV Cache Size=2562026.0390.97
SnapKVLLM=Qwen3-8B, KV Cache Size=2562026.0390.41
PyramidKVLLM=Qwen3-8B, KV Cache Size=2562026.0388.5
Transformer2026.0487
DapQLLM=Qwen2.5-7B-Instruct, KV Cache Size=2562026.0385.11
DapQLLM=Llama3.1-8B-Instruct, KV Cache Size=2562026.0384.7
SnapKVLLM=Qwen2.5-7B-Instruct, KV Cache Size=2562026.0384.3
PyramidKVLLM=Qwen2.5-7B-Instruct, KV Cache Size=2562026.0383.8
PyramidKVLLM=Llama3-8B-Instruct, KV Cache Size=1282026.0379.67
PyramidKVLLM=Llama3.1-8B-Instruct, KV Cache Size=2562026.0378.3
DapQLLM=Qwen3-8B, KV Cache Size=1282026.0377.89
DapQLLM=Qwen2.5-7B-Instruct, KV Cache Size=1282026.0376.25
DapQLLM=Qwen2.5-7B-Instruct, KV Cache Size=642026.0375.75
H2OLLM=Qwen2.5-7B-Instruct, KV Cache Size=2562026.0375.64
SnapKVLLM=Qwen3-8B, KV Cache Size=1282026.0375.39
SnapKVLLM=Llama3.1-8B-Instruct, KV Cache Size=2562026.0374.84
PyramidKVLLM=Qwen2.5-7B-Instruct, KV Cache Size=1282026.0374.8
SnapKVLLM=Llama3-8B-Instruct, KV Cache Size=1282026.0374.67
H2OLLM=Qwen3-8B, KV Cache Size=2562026.0374.55
SnapKVLLM=Qwen2.5-7B-Instruct, KV Cache Size=1282026.0373.64
SnapKVLLM=Qwen2.5-7B-Instruct, KV Cache Size=642026.0372.84
PyramidKVLLM=Qwen3-8B, KV Cache Size=1282026.0372.36
H2OLLM=Qwen2.5-7B-Instruct, KV Cache Size=1282026.0370.45
DapQLLM=Llama3.1-8B-Instruct, KV Cache Size=1282026.0370.34
DapQLLM=Llama3-8B-Instruct, KV Cache Size=642026.0368.34
H2OLLM=Qwen3-8B, KV Cache Size=1282026.0367.5
H2OLLM=Llama3-8B-Instruct, KV Cache Size=2562026.0366.81
PyramidKVLLM=Llama3.1-8B-Instruct, KV Cache Size=1282026.0365.23
H2OLLM=Qwen2.5-7B-Instruct, KV Cache Size=642026.0363.7
H2OLLM=Qwen3-8B, KV Cache Size=642026.0362.7
DapQLLM=Llama3.1-8B-Instruct, KV Cache Size=642026.0362.2
DapQLLM=Qwen3-8B, KV Cache Size=642026.0361.98
SnapKVLLM=Llama3-8B-Instruct, KV Cache Size=642026.0361.7
SnapKVLLM=Llama3.1-8B-Instruct, KV Cache Size=1282026.0361.45
PyramidKVLLM=Qwen3-8B, KV Cache Size=642026.0361.32
H2OLLM=Llama3.1-8B-Instruct, KV Cache Size=2562026.0361.18
SnapKVLLM=Qwen3-8B, KV Cache Size=642026.0359.73
SnapKVLLM=Llama3.1-8B-Instruct, KV Cache Size=642026.0356.5
PyramidKVLLM=Qwen2.5-7B-Instruct, KV Cache Size=642026.0356.11
PyramidKVLLM=Llama3-8B-Instruct, KV Cache Size=642026.0355.45
PyramidKVLLM=Llama3.1-8B-Instruct, KV Cache Size=642026.0352.25
H2OLLM=Llama3-8B-Instruct, KV Cache Size=1282026.0350.92
H2OLLM=Llama3.1-8B-Instruct, KV Cache Size=1282026.0347.61
H2OLLM=Llama3-8B-Instruct, KV Cache Size=642026.0342.37
H2OLLM=Llama3.1-8B-Instruct, KV Cache Size=642026.0340.36