List your company

Language Modeling on PG-19

5PerplexityCREAM
4.835.97757.1258.2725Jun 11, 2024Oct 10, 2024Feb 9, 2025Jun 11, 2025Oct 11, 2025Feb 10, 2026Jun 12, 2026

Evaluation results

MethodLinks
CREAMBackbone=LLaMa2-7B, Interpolation Method=NTK, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=64K2024.065——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=YaRN, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=64K2024.065——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=NTK, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=96K2024.065.1——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=NTK, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=64K2024.065.1——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=NTK, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=128K2024.065.2——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=NTK, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=192K2024.065.2——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=YaRN, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=96K2024.065.2——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=YaRN, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=128K2024.065.2——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=YaRN, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=64K2024.065.2——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=YaRN, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=192K2024.065.3——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=NTK, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=96K2024.065.3——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=NTK, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=128K2024.065.3——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=YaRN, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=96K2024.065.3——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=NTK, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=192K2024.065.4——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=NTK, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=256K2024.065.4——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=YaRN, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=128K2024.065.4——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=YaRN, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=192K2024.065.4——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=YaRN, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=256K2024.065.5——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=Linear, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=64K2024.065.9——————
SHAREDLLMContext Length=128K2026.035.96——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=Linear, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=96K2024.066——————
LLaMA-3.1Context Length=128K2026.036.03——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=Linear, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=128K2024.066.1——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=Linear, Fine-tuning Context Length=4K, Target Extrapolation Length=192K, Evaluation Context Length=192K2024.066.1——————
CEPEContext Length=128K2026.036.1——————
SHAREDLLMContext Length=8K2026.036.28——————
CEPEContext Length=8K2026.036.4——————
LLaMA-3.1Context Length=8K2026.036.52——————
SHAREDLLMContext Length=4K2026.036.55——————
YaRNContext Length=8K2026.036.62——————
SHAREDLLMContext Length=32K2026.036.65——————
CEPEContext Length=4K2026.036.69——————
PIContext Length=8K2026.036.77——————
LLaMA-3.1Context Length=4K2026.036.77——————
CEPEContext Length=32K2026.036.8——————
LLaMA-2-32KContext Length=8K2026.036.81——————
Full KVcontext length=4k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=Full2026.056.81——————
LLaMA-3.1Context Length=32K2026.036.84——————
YaRNContext Length=4K2026.036.85——————
PIContext Length=32K2026.036.89——————
YaRNContext Length=32K2026.036.91——————
LLaMA-2-32KContext Length=4K2026.036.93——————
PIContext Length=4K2026.036.97——————
LLaMA-2-32KContext Length=32K2026.037.04——————
Meta-Softcontext length=4k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=2562026.057.05——————
Judge Qcontext length=4k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=2562026.057.11——————
ZeroMergecontext length=4k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=2562026.057.13——————
YARN-128k-7bModel Size=7B, Context Length=4k, GPU Configuration=Single GPU2025.037.17——————
YARN-128k-7BContext Length=4k, Model Size=7B2026.067.17——————
YARN-128k-7bModel Size=7B, Context Length=1k, GPU Configuration=Single GPU2025.037.22——————
YARN-128k-7BContext Length=1k, Model Size=7B2026.067.22——————
Full KVcontext length=16k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=Full2026.057.23——————
SnapKVcontext length=4k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=2562026.057.24——————
H2Ocontext length=4k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=2562026.057.27——————
YARN-128k-7bModel Size=7B, Context Length=2k, GPU Configuration=Single GPU2025.037.47——————
YARN-128k-7BContext Length=2k, Model Size=7B2026.067.47——————
Meta-Softcontext length=16k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=2562026.057.49——————
Judge Qcontext length=16k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=2562026.057.58——————
LLaMA-2 70BModel=LLaMA-2 70B2026.037.6——————
LongLoRA-7B-32kModel Size=7B, Context Length=16k, GPU Configuration=Single GPU2025.037.62——————
LongLoRA-7B-32kContext Length=16k, Model Size=7B2026.067.62——————
ZeroMergecontext length=16k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=2562026.057.63——————
SnapKVcontext length=16k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=2562026.057.72——————
H2Ocontext length=16k, Backbone=Llama-3.1-8B-Instruct, KV cache compression (B)=2562026.057.79——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=Linear, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=64K2024.067.8——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=Linear, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=96K2024.068——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=Linear, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=128K2024.068——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=Linear, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=192K2024.068.1——————
CREAMBackbone=LLaMa2-7B, Interpolation Method=Linear, Fine-tuning Context Length=4K, Target Extrapolation Length=256K, Evaluation Context Length=256K2024.068.2——————
Focus (LLaMA-2 70B)Backbone=LLaMA-2 70B, Centroid params=10.5M (0.015%), K=4, dg=16, top-k=22026.038.3——————
SelfExtendBackbone=Llama2-7b-chat-4k, Context Window=1k2025.028.81——————
Dyn-NTKBackbone=Llama2-7b-chat-4k, Context Window=1k2025.028.81——————
GALIBackbone=Llama2-7b-chat-4k, Context Window=1k2025.028.81——————
LLaMA-2-7BModel Size=7B, Context Length=4k, GPU Configuration=Single GPU2025.038.92——————
LLaMA-2-7BContext Length=4k, Model Size=7B2026.068.92——————
NTKBackbone=Llama2-7b-chat-4k, Context Window=1k2025.028.95——————
YaRNBackbone=Llama2-7b-chat-4k, Context Window=2k2025.028.97——————
SelfExtendBackbone=Llama2-7b-chat-4k, Context Window=2k2025.028.99——————
Dyn-NTKBackbone=Llama2-7b-chat-4k, Context Window=2k2025.028.99——————
GALIBackbone=Llama2-7b-chat-4k, Context Window=2k2025.028.99——————
YaRNBackbone=Llama2-7b-chat-4k, Context Window=7k2025.029.03——————
NTKBackbone=Llama2-7b-chat-4k, Context Window=2k2025.029.04——————
YaRNBackbone=Llama2-7b-chat-4k, Context Window=8k2025.029.04——————
NTKBackbone=Llama2-7b-chat-4k, Context Window=6k2025.029.06——————
ChunkLlamaBackbone=Llama2-7b-chat-4k, Context Window=1k2025.029.07——————
YaRNBackbone=Llama2-7b-chat-4k, Context Window=3k2025.029.07——————
Dyn-NTKBackbone=Llama2-7b-chat-4k, Context Window=3k2025.029.15——————
YaRNBackbone=Llama2-7b-chat-4k, Context Window=6k2025.029.15——————
GALIBackbone=Llama2-7b-chat-4k, Context Window=3k2025.029.15——————
SelfExtendBackbone=Llama2-7b-chat-4k, Context Window=3k2025.029.16——————
SelfExtendBackbone=Llama2-7b-chat-4k, Context Window=6k2025.029.16——————
NTKBackbone=Llama2-7b-chat-4k, Context Window=3k2025.029.16——————
NTKBackbone=Llama2-7b-chat-4k, Context Window=5k2025.029.16——————
YaRNBackbone=Llama2-7b-chat-4k, Context Window=4k2025.029.16——————
YaRNBackbone=Llama2-7b-chat-4k, Context Window=5k2025.029.17——————
NTKBackbone=Llama2-7b-chat-4k, Context Window=4k2025.029.18——————
SelfExtendBackbone=Llama2-7b-chat-4k, Context Window=7k2025.029.2——————
SelfExtendBackbone=Llama2-7b-chat-4k, Context Window=4k2025.029.24——————
GALIBackbone=Llama2-7b-chat-4k, Context Window=4k2025.029.24——————
SelfExtendBackbone=Llama2-7b-chat-4k, Context Window=5k2025.029.25——————