List your company

Language Modeling on FineWeb-Edu 500M-token (val)

2.221Valid LossFusedKV
2.20282.325652.44852.57135Dec 3, 2025

Evaluation results

MethodLinks
FusedKVModel Scale=1.5B2025.122.221
FusedKV-LiteModel Scale=1.5B2025.122.225
VanillaModel Scale=1.5B, Cache Mem.=12025.122.241
YOCOModel Scale=1.5B2025.122.244
GQAModel Scale=1.5B2025.122.245
CLAModel Scale=1.5B, Cache Mem.=1/22025.122.258
FusedKV-LiteModel Scale=650M2025.122.473
FusedKVModel Scale=650M2025.122.474
VanillaModel Scale=650M, Cache Mem.=12025.122.483
GQAModel Scale=650M2025.122.497
YOCOModel Scale=650M2025.122.498
CLAModel Scale=650M, Cache Mem.=1/22025.122.511
FusedKV-LiteModel Scale=332M2025.122.639
FusedKVModel Scale=332M2025.122.642
VanillaModel Scale=332M, Cache Mem.=12025.122.651
GQAModel Scale=332M2025.122.662
YOCOModel Scale=332M2025.122.663
CLAModel Scale=332M, Cache Mem.=1/22025.122.676