List your company

Long-context language understanding on LongBench (Aggregate Score Only)

49.8Aggregate ScoreFullKV
24.299230.919637.5444.1604Jun 23, 2026

Evaluation results

MethodLinks
FullKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=Full2026.0649.8
FullKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=Full2026.0649.08
CompressKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=10242026.0648.69
FullKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=Full2026.0648.57
CAKELLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=10242026.0648.52
HeadKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=10242026.0648.42
CompressKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=10242026.0648.24
AdaKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=10242026.0648.19
SnapKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=10242026.0648.18
CAKELLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=10242026.0647.97
AdaKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=10242026.0647.94
FullKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=Full2026.0647.82
SnapKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=10242026.0647.82
CompressKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=10242026.0647.78
PyramidKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=10242026.0647.7
PyramidKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=10242026.0647.65
CAKELLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=10242026.0647.51
HeadKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=10242026.0647.5
SnapKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=10242026.0647.34
AdaKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=10242026.0647.33
HeadKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=10242026.0647.05
PyramidKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=10242026.0646.98
CompressKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=10242026.0646.96
CompressKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=2562026.0646.71
CAKELLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=10242026.0646.66
SnapKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=10242026.0646.48
HeadKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=10242026.0646.41
AdaKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=10242026.0646.38
CAKELLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=2562026.0646.3
PyramidKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=10242026.0645.96
CompressKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=2562026.0645.43
CompressKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=2562026.0645.37
SnapKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=2562026.0645.21
CAKELLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=2562026.0644.73
CompressKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=2562026.0644.73
CAKELLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=2562026.0644.7
CAKELLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=2562026.0644.49
AdaKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=2562026.0644.45
PyramidKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=2562026.0644.36
HeadKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=2562026.0644.21
HeadKVLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=2562026.0644.11
HeadKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=2562026.0644.1
HeadKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=2562026.0644.02
SnapKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=2562026.0643.77
SnapKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=2562026.0643.76
AdaKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=2562026.0643.75
AdaKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=2562026.0643.68
SnapKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=2562026.0643.36
AdaKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=2562026.0643.3
PyramidKVLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=2562026.0643.06
PyramidKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=2562026.0642.71
PyramidKVLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=2562026.0642.11
StreamingLLMLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=10242026.0636.95
StreamingLLMLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=10242026.0634.73
StreamingLLMLLM Scale=Small-scale LLMs, Backbone=Llama-3.1-8B, KV-cache budget=2562026.0633.92
StreamingLLMLLM Scale=Small-scale LLMs, Backbone=Mistral-7B, KV-cache budget=2562026.0631.22
StreamingLLMLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=10242026.0629.84
StreamingLLMLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=10242026.0629.62
StreamingLLMLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-14B, KV-cache budget=2562026.0625.86
StreamingLLMLLM Scale=Large-scale LLMs, Backbone=Qwen2.5-32B, KV-cache budget=2562026.0625.28