List your company

Mathematical Reasoning on MATH 18 (test)

39.7m1@t1GPT-3.5
0.38810.59420.831.006Jul 25, 2024

Evaluation results

MethodLinks
GPT-3.52024.0739.747.845.154.3——
Self-RefineBackbone=GPT-3.52024.0739.746.5——36.546.7
Eurus-7b-SFT2024.0712.319.816.322.9——
Self-RefineBackbone=Eurus-7b-SFT2024.0712.319.4——915.1
RISEBackbone=Mistral-7B, Iteration=Base2024.077.5138.413——
Mistral-7B2024.077.5138.413——
Self-RefineBackbone=Mistral-7B2024.077.512.3——7.111.4
RISEBackbone=Mistral-7B, Iteration=+Iteration 12024.076.79.518.429.7——
RISEBackbone=Llama2 Base, Iteration=+Iteration 12024.076.38.89.719.4——
Self-RefineBackbone=Llama2 Base, Iteration=+Iteration 12024.076.38.7——5.99.9
RISEBackbone=Llama2 Base, Iteration=+Iteration 22024.075.810.410.419.8——
SFT on oracle dataData Type=Only correct data2024.075.87.95.512.1——
Self-RefineBackbone=Llama2 Base, Iteration=+Iteration 22024.075.89.4——5.79.5
RISEBackbone=Llama2 Base, Iteration=+Boost2024.075.56.85.514.6——
Self-RefineBackbone=Llama2 Base, Iteration=+Boost2024.075.56.5——2.97.2
SFT on oracle dataData Type=Correct and incorrect2024.0755.2513.1——
RISEBackbone=Llama2 Base, Iteration=Base2024.071.95.11.42.3——
Self-RefineBackbone=Llama2 Base, Iteration=Base2024.071.95.1——1.93.1