List your company

Multi-Document Question Answering on LongBench

57.67HotpotQA AccFullKV
13.137224.698636.2647.8214Feb 9, 2026Feb 14, 2026Feb 19, 2026Feb 24, 2026Mar 1, 2026Mar 6, 2026Mar 12, 2026

Evaluation results

MethodLinks
FullKVBackbone=Qwen3-8B, KV Cache Size=Full2026.0357.6744.73——
FullKVBackbone=Qwen2.5-7B-Instruct, KV Cache Size=Full2026.0355.9144.7——
DapQBackbone=Qwen3-8B, KV Cache Size=2562026.0354.7944.47——
SnapKVBackbone=Qwen3-8B, KV Cache Size=2562026.0354.3841.4——
DapQBackbone=Qwen3-8B, KV Cache Size=1282026.0353.8443——
DapQBackbone=Qwen2.5-7B-Instruct, KV Cache Size=2562026.0351.9241.46——
SnapKVBackbone=Qwen2.5-7B-Instruct, KV Cache Size=2562026.0351.7640.78——
H2OBackbone=Qwen3-8B, KV Cache Size=2562026.0351.3740.5——
SnapKVBackbone=Qwen3-8B, KV Cache Size=1282026.0351.241.66——
PyramidKVBackbone=Qwen3-8B, KV Cache Size=2562026.035141.37——
DapQBackbone=Llama3-8B-Instruct, KV Cache Size=2562026.0350.6734.35——
PyramidKVBackbone=Llama3-8B-Instruct, KV Cache Size=2562026.0350.233.88——
FullKVBackbone=Llama3-8B-Instruct, KV Cache Size=Full2026.0350.1734.88——
DapQBackbone=Llama3-8B-Instruct, KV Cache Size=1282026.0350.0433.59——
PyramidKVBackbone=Qwen2.5-7B-Instruct, KV Cache Size=2562026.0349.8940.77——
DapQBackbone=Qwen3-8B, KV Cache Size=642026.0349.8441.28——
SnapKVBackbone=Llama3-8B-Instruct, KV Cache Size=2562026.0349.7533.8——
DapQBackbone=Qwen2.5-7B-Instruct, KV Cache Size=642026.0349.6641.25——
DapQBackbone=Qwen2.5-7B-Instruct, KV Cache Size=1282026.0349.6241.36——
PyramidKVBackbone=Llama3-8B-Instruct, KV Cache Size=1282026.0349.5231.6——
SnapKVBackbone=Llama3-8B-Instruct, KV Cache Size=1282026.0349.3632.4——
DapQBackbone=Llama3-8B-Instruct, KV Cache Size=642026.0349.1132.88——
H2OBackbone=Llama3-8B-Instruct, KV Cache Size=1282026.0348.6531.9——
H2OBackbone=Llama3-8B-Instruct, KV Cache Size=2562026.0348.6231.5——
SnapKVBackbone=Qwen2.5-7B-Instruct, KV Cache Size=1282026.0348.6241.4——
PyramidKVBackbone=Qwen3-8B, KV Cache Size=1282026.0348.4139.46——
H2OBackbone=Llama3-8B-Instruct, KV Cache Size=642026.0348.2731.7——
H2OBackbone=Qwen3-8B, KV Cache Size=1282026.0348.139.85——
H2OBackbone=Qwen2.5-7B-Instruct, KV Cache Size=2562026.0347.8940.97——
SnapKVBackbone=Llama3-8B-Instruct, KV Cache Size=642026.0347.1631.71——
PyramidKVBackbone=Llama3-8B-Instruct, KV Cache Size=642026.0347.0131.54——
H2OBackbone=Qwen3-8B, KV Cache Size=642026.0346.6639.27——
PyramidKVBackbone=Qwen3-8B, KV Cache Size=642026.0346.6139.2——
SnapKVBackbone=Qwen3-8B, KV Cache Size=642026.0346.5839.38——
PyramidKVBackbone=Qwen2.5-7B-Instruct, KV Cache Size=1282026.0346.5739.07——
SnapKVBackbone=Qwen2.5-7B-Instruct, KV Cache Size=642026.0345.5640.28——
H2OBackbone=Qwen2.5-7B-Instruct, KV Cache Size=1282026.0345.1439.77——
H2OBackbone=Qwen2.5-7B-Instruct, KV Cache Size=642026.0344.9439.06——
PyramidKVBackbone=Qwen2.5-7B-Instruct, KV Cache Size=642026.0343.5138.08——
FullKVModel=LLaMA-3-8B-Instruct, KV cache size=Full2026.0243.0936.1521.77—
CompilerKVModel=LLaMA-3-8B-Instruct, KV cache size=1282026.0241.3728.4821.25—
StreamingLLMModel=LLaMA-3-8B-Instruct, KV cache size=1282026.0237.8429.0216.77—
CompilerKVModel=Mistral-7B-Instruct-v0.2, KV cache size=1282026.0236.4222.9116.47—
CompilerKVModel=InternLM-2.5-Chat-1M, KV cache size=1282026.0232.1719.7520.95—
CompilerKVModel=Qwen2-7B-Instruct, KV cache size=1282026.0214.8516.458.4—
baselinePrecision=16-bit, Backbone=Llama3.1-8B2025.03———39.96
KiViPrecision=2-bit, Backbone=Llama3.1-8B2025.03———34.02
KiViPrecision=4-bit, Backbone=Llama3.1-8B2025.03———39.74
LogQuantPrecision=2-bit, Backbone=Llama3.1-8B2025.03———36.08
LogQuantPrecision=4-bit, Backbone=Llama3.1-8B2025.03———39.93