List your company

Natural Language Reasoning on DROP

89.62AccuracyInfiGFusion
48.207258.958669.7180.4614Feb 18, 2025Mar 5, 2025Mar 20, 2025Apr 4, 2025Apr 19, 2025May 4, 2025May 20, 2025

Evaluation results

MethodLinks
InfiGFusionModel Size=14B, GPU Hour=1952025.0589.62
Pivot-SFTModel Size=14B, GPU Hour=1202025.0589.44
InfiFusionModel Size=14B, GPU Hour=1602025.0589.27
FuseChatModel Size=14B, GPU Hour=6502025.0589.23
GPT-4o + QuaSARBackbone=GPT-4o, Prompting Strategy=QuaSAR2025.0288.9
FuseLLMModel Size=14B, GPU Hour=2252025.0588.74
Phi-4Model Size=14B, GPU Hour=∼1.0M2025.0588.67
MiniLogitModel Size=14B, GPU Hour=2202025.0588.56
Llama-3-70B + QuaSARBackbone=Llama-3-70B, Prompting Strategy=QuaSAR2025.0288
Mistral-SmallModel Size=24B, GPU Hour=∼1.6M2025.0586.52
Qwen2.5-InstructModel Size=14B, GPU Hour=∼1.8M2025.0585.56
Qwen2.5-CoderModel Size=14B, GPU Hour=∼1.8M2025.0584.34
GPT-4o + CoTBackbone=GPT-4o, Prompting Strategy=CoT2025.0284.2
GPT-4oBackbone=GPT-4o, Prompting Strategy=Zero-shot2025.0283.4
Llama-3-70BBackbone=Llama-3-70B, Prompting Strategy=Zero-shot2025.0281.4
Llama-3-70B + CoTBackbone=Llama-3-70B, Prompting Strategy=CoT2025.0280.2
Qwen2-72B + QuaSARBackbone=Qwen2-72B, Prompting Strategy=QuaSAR2025.0269
Qwen2-72BBackbone=Qwen2-72B, Prompting Strategy=Zero-shot2025.0266.4
Qwen2-72B + CoTBackbone=Qwen2-72B, Prompting Strategy=CoT2025.0264
QuaSARBackbone=Llama-3-8B, Protocol=Annotation Tuning2025.0263.9
Llama-3-8BStrategy=Base, Protocol=Annotation Tuning2025.0260.2
QuaSARBackbone=Qwen2-7B, Protocol=Annotation Tuning2025.0260
COTBackbone=Llama-3-8B, Protocol=Annotation Tuning2025.0259.3
QuaSARBackbone=Llama-3-8B, Protocol=ICL2025.0258.7
Llama-3-8BStrategy=Base, Protocol=ICL2025.0258.4
COTBackbone=Llama-3-8B, Protocol=ICL2025.0257.9
QuaSARBackbone=Qwen2-1.5B, Protocol=Annotation Tuning2025.0257.3
QuaSARBackbone=Llama-3-1B, Protocol=Annotation Tuning2025.0257.2
COTBackbone=Qwen2-7B, Protocol=Annotation Tuning2025.0256.8
COTBackbone=Qwen2-7B, Protocol=ICL2025.0256
Qwen2-7BStrategy=Base, Protocol=ICL2025.0255.3
COTBackbone=Llama-3-1B, Protocol=Annotation Tuning2025.0255
QuaSARBackbone=Qwen2-7B, Protocol=ICL2025.0254.6
COTBackbone=Llama-3-1B, Protocol=ICL2025.0254.4
Qwen2-7BStrategy=Base, Protocol=Annotation Tuning2025.0254.2
QuaSARBackbone=Llama-3-1B, Protocol=ICL2025.0254.1
Llama-3-1BStrategy=Base, Protocol=Annotation Tuning2025.0253
Llama-3-1BStrategy=Base, Protocol=ICL2025.0252.5
COTBackbone=Qwen2-1.5B, Protocol=Annotation Tuning2025.0252.5
COTBackbone=Qwen2-1.5B, Protocol=ICL2025.0251.2
Qwen2-1.5BStrategy=Base, Protocol=Annotation Tuning2025.0250.8
QuaSARBackbone=Qwen2-1.5B, Protocol=ICL2025.0250.8
Qwen2-1.5BStrategy=Base, Protocol=ICL2025.0249.8