Edit

Multilingual exam, understanding, mathematics, and translation evaluations.

multilingual_understanding vs. estimated ISO FLOPs

9 matching runs · 6ND; missing tokens estimated at 20× parameters
All 10 filtered runs
Run Scale (B) ISO FLOPs Training datamultilingual_understanding multilingual_exam multilingual_mathematics multilingual_translation
literature/qwen2_72b
(unverified)
~72.000~6.2e2380.7076.6076.0037.80
literature/llama_3_70b
(unverified)
~70.000~5.9e2379.9070.0067.1038.00
literature/qwen1.5_110b
(unverified)
~110.000~1.5e2478.2075.6064.4036.20
literature/qwen1.5_72b
(unverified)
~72.000~6.2e2378.2066.4061.7035.60
literature/mixtral_8x22b
(unverified)
Unknown77.7063.5062.9023.30
literature/qwen2_7b
(unverified)
~7.000~5.9e2172.0059.2057.5031.50
literature/llama_3_8b
(unverified)
~8.000~7.7e2168.6052.3036.3031.90
literature/qwen1.5_7b
(unverified)
~7.000~5.9e2167.6047.7037.3028.40
literature/mistral_7b_v0.2
(unverified)
~7.000~5.9e2163.3047.1026.3023.30
literature/gemma_7b
(unverified)
~7.000~5.9e2158.3042.7039.1031.20
10 models