List your company

Physics problem solving on Scibench fund

74.6AccuracyPoT + REFTOOL
35.0845.3455.665.86May 27, 2025

Evaluation results

MethodLinks
PoT + REFTOOLBackbone=GPT-4o2025.0574.6—
PoTBackbone=GPT-4o2025.0573.2—
PoT + RAGBackbone=Gemini-1.5-Pro2025.0573.2—
PoT + RAGBackbone=GPT-4o2025.0573.2—
PoT + REFTOOLBackbone=Gemini-1.5-Pro2025.0573.2—
Physics ReasonerBackbone=Gemini-1.5-Pro, Few-shot=4-shot2025.0573.2—
Physics ReasonerBackbone=GPT-4o, Few-shot=4-shot2025.0571.8—
CreatorBackbone=Gemini-1.5-Pro2025.0570.4—
PoTBackbone=Gemini-1.5-Pro2025.0569—
PoT + REFTOOLBackbone=GPT-42025.0564.8—
Physics ReasonerBackbone=GPT-42025.0563.4—
Physics ReasonerBackbone=GPT-4, Few-shot=4-shot2025.0563.4—
Physics ReasonerBackbone=Llama-3.1-70B, Few-shot=4-shot2025.0562—
Physics ReasonerBackbone=Gemini-1.5-Pro2025.0559.2—
PoTBackbone=GPT-42025.0559.2—
PoT + RAGBackbone=GPT-42025.0559.2—
CreatorBackbone=GPT-4o2025.0557.7—
PoT + REFTOOLBackbone=Llama-3.1-70B2025.0557.7—
Physics ReasonerBackbone=Llama-3.1-70B2025.0556.3—
CreatorBackbone=Llama-3.1-70B2025.0556.3—
PoT + RAGBackbone=Llama-3.1-70B2025.0554.9—
CreatorBackbone=GPT-42025.0553.5—
PoTBackbone=Llama-3.1-70B2025.0553.5—
Physics ReasonerBackbone=GPT-4o2025.0536.6—
Creator2025.05—59.5
Physics Reasoner2025.05—53.9
Physics ReasonerFew-shot=4-shot2025.05—67.6
PoT2025.05—63.7
PoT + RAG2025.05—65.1
PoT + REFTOOL2025.05—67.6