Compositional Generalization on Evaluation Dataset (Fold 3 Seen)
66.69ScoreLLaMA-3-8B + COGLM
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaMA-3-8B + COGLMFine-tuning Strategy=LoRA, Model Backbone=LLaMA-3-8B, Method Components=COGLM2026.01 | 66.69 | |
| Mistral-7B + COGLMFine-tuning Strategy=LoRA, Model Backbone=Mistral-7B, Method Components=COGLM2026.01 | 65.1 | |
| Mistral-7BFine-tuning Strategy=LoRA, Model Backbone=Mistral-7B2026.01 | 62.64 | |
| Qwen2.5-7B + COGLMFine-tuning Strategy=LoRA, Model Backbone=Qwen2.5-7B, Method Components=COGLM2026.01 | 61.73 | |
| Gemma-7B + COGLMFine-tuning Strategy=LoRA, Model Backbone=Gemma-7B, Method Components=COGLM2026.01 | 61.7 | |
| Qwen2.5-7BFine-tuning Strategy=LoRA, Model Backbone=Qwen2.5-7B2026.01 | 61.09 | |
| Gemma-7BFine-tuning Strategy=LoRA, Model Backbone=Gemma-7B2026.01 | 60.43 | |
| LLaMA-3-8BFine-tuning Strategy=LoRA, Model Backbone=LLaMA-3-8B2026.01 | 59.72 | |
| DeepSeek V3 + FS*Evaluation Protocol=Few-shot, Model Backbone=DeepSeek V32026.01 | 46.62 | |
| DeepSeek-7B + COGLMFine-tuning Strategy=LoRA, Model Backbone=DeepSeek-7B, Method Components=COGLM2026.01 | 45.47 | |
| Claude 3.5 Sonnet + FS*Evaluation Protocol=Few-shot, Model Backbone=Claude 3.5 Sonnet2026.01 | 45.04 | |
| DeepSeek V3Evaluation Protocol=Zero-shot, Model Backbone=DeepSeek V32026.01 | 42.46 | |
| Claude 3.5 SonnetEvaluation Protocol=Zero-shot, Model Backbone=Claude 3.5 Sonnet2026.01 | 41.59 | |
| GPT-4o + FS*Evaluation Protocol=Few-shot, Model Backbone=GPT-4o2026.01 | 40.05 | |
| DeepSeek-7BFine-tuning Strategy=LoRA, Model Backbone=DeepSeek-7B2026.01 | 39.02 | |
| GPT-4oEvaluation Protocol=Zero-shot, Model Backbone=GPT-4o2026.01 | 35.58 | |
| Llama 3 (70B) + FS*Evaluation Protocol=Few-shot, Model Backbone=Llama 3 (70B)2026.01 | 33.15 | |
| Llama 3 (70B)Evaluation Protocol=Zero-shot, Model Backbone=Llama 3 (70B)2026.01 | 22.31 |