List your company

Long-context reasoning on LongGenBench 8K

44.51GSM8K ScoreFull Cache
3.086813.840924.59535.3491May 28, 2026

Evaluation results

MethodLinks
Full CacheBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=Full Cache2026.0544.5150.9264.8153.41
SCOPEBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=1024 Tokens2026.0544.24966.1653.12
Moment-KVBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=1024 Tokens2026.0540.315066.3352.21
StreamingLLMBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=1024 Tokens2026.0539.3648.2963.8350.49
H2OBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=1024 Tokens2026.0538.2948.0264.950.4
Moment-KVBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=512 Tokens2026.0535.6349.9263.3349.63
PyramidInferBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=1024 Tokens2026.0534.284557.1645.48
StreamingLLMBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=512 Tokens2026.0533.8848.5364.2548.89
SCOPEBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=512 Tokens2026.0527.8547.3963.2546.16
H2OBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=512 Tokens2026.0525.0744.3847.1638.87
PyramidInferBase Model=LLaMA-3.1-8B-Instruct, Decoding Budget=512 Tokens2026.0524.6842.725239.8
Full CacheBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=Full Cache2026.059.4520.2851.6327.12
Moment-KVBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=1024 Tokens2026.058.819.6353.3327.25
Moment-KVBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=512 Tokens2026.057.6119.6250.0825.77
PyramidInferBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=1024 Tokens2026.057.2219.5142.9123.21
StreamingLLMBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=1024 Tokens2026.056.5821.0846.524.72
SCOPEBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=512 Tokens2026.056.5816.4451.7524.92
H2OBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=1024 Tokens2026.056.0320.2248.4124.89
PyramidInferBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=512 Tokens2026.055.8717.8639.8321.19
SCOPEBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=1024 Tokens2026.055.791853.2525.68
H2OBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=512 Tokens2026.055.6318.3143.2522.4
StreamingLLMBase Model=Mistral-7B-Instruct-v0.3, Decoding Budget=512 Tokens2026.054.6818.6453.5825.63