List your company
ResearchFormerly WizWand

Compositional Generalization on Evaluation Dataset Unseen (Fold 1)

0.4818ScoreDeepSeek V3
0.2432240.3051620.36710.429038Jan 29, 2026
Updated 8mo ago

Evaluation results

MethodLinks
DeepSeek V3Evaluation Protocol=Zero-shot, Model Backbone=DeepSeek V32026.010.4818
GPT-4o + FS*Evaluation Protocol=Few-shot, Model Backbone=GPT-4o2026.010.4668
Claude 3.5 Sonnet + FS*Evaluation Protocol=Few-shot, Model Backbone=Claude 3.5 Sonnet2026.010.4627
LLaMA-3-8BFine-tuning Strategy=LoRA, Model Backbone=LLaMA-3-8B2026.010.4626
Gemma-7BFine-tuning Strategy=LoRA, Model Backbone=Gemma-7B2026.010.451
GPT-4oEvaluation Protocol=Zero-shot, Model Backbone=GPT-4o2026.010.4298
LLaMA-3-8B + COGLMFine-tuning Strategy=LoRA, Model Backbone=LLaMA-3-8B, Method Components=COGLM2026.010.4109
Qwen2.5-7BFine-tuning Strategy=LoRA, Model Backbone=Qwen2.5-7B2026.010.4079
Gemma-7B + COGLMFine-tuning Strategy=LoRA, Model Backbone=Gemma-7B, Method Components=COGLM2026.010.3974
Claude 3.5 SonnetEvaluation Protocol=Zero-shot, Model Backbone=Claude 3.5 Sonnet2026.010.3944
Mistral-7BFine-tuning Strategy=LoRA, Model Backbone=Mistral-7B2026.010.388
Llama 3 (70B) + FS*Evaluation Protocol=Few-shot, Model Backbone=Llama 3 (70B)2026.010.3811
Mistral-7B + COGLMFine-tuning Strategy=LoRA, Model Backbone=Mistral-7B, Method Components=COGLM2026.010.3786
Qwen2.5-7B + COGLMFine-tuning Strategy=LoRA, Model Backbone=Qwen2.5-7B, Method Components=COGLM2026.010.344
DeepSeek-7BFine-tuning Strategy=LoRA, Model Backbone=DeepSeek-7B2026.010.3082
DeepSeek V3 + FS*Evaluation Protocol=Few-shot, Model Backbone=DeepSeek V32026.010.3016
DeepSeek-7B + COGLMFine-tuning Strategy=LoRA, Model Backbone=DeepSeek-7B, Method Components=COGLM2026.010.2953
Llama 3 (70B)Evaluation Protocol=Zero-shot, Model Backbone=Llama 3 (70B)2026.010.2524