Coreference Resolution on CLUEWSC
90.98EMC-DPO
Evaluation results
| Method | Links | |
|---|---|---|
| C-DPOSize=70B, Evaluation Protocol=zero-shot2024.09 | 90.98 | |
| C-SFTSize=70B, Evaluation Protocol=zero-shot2024.09 | 90.88 | |
| Llama-3 InstructSize=70B, Evaluation Protocol=zero-shot2024.09 | 86.37 | |
| C-DPOSize=8B, Evaluation Protocol=zero-shot2024.09 | 85.04 | |
| GLM-5 BaseArchitecture=MoE, Activated Params=40B, Total Params=744B2026.02 | 84.2 | |
| GLM-4.5 BaseArchitecture=MoE, Activated Params=32B, Total Params=355B2026.02 | 83.5 | |
| DeepSeek-V3 BaseArchitecture=MoE, Activated Params=37B, Total Params=671B2026.02 | 82.7 | |
| C-SFTSize=8B, Evaluation Protocol=zero-shot2024.09 | 80.64 | |
| Llama-3 InstructSize=8B, Evaluation Protocol=zero-shot2024.09 | 80.12 | |
| DeepSeek 7B (Dense)# Shot=5-shot, # Total Params=6.9B, # Activated Params=6.9B, FLOPs per 4K Tokens=183.5T, # Training Tokens=2T2024.01 | 73.1 | |
| DeepSeekMoE 142B (Half Activated)# Shot=5-shot2024.01 | 72.6 | |
| DeepSeekMoE 16B# Shot=5-shot, # Total Params=16.4B, # Activated Params=2.8B, FLOPs per 4K Tokens=74.4T, # Training Tokens=2T2024.01 | 72.1 | |
| DeepSeekMoE 16B# Shot=5-shot, # Total Params=16.4B, # Activated Params=2.8B, FLOPs per 4K Tokens=74.4T, # Training Tokens=2T2024.01 | 72.1 | |
| DeepSeekMoE 145B# Shot=5-shot2024.01 | 71.9 | |
| DeepSeek 67B (Dense)# Shot=5-shot2024.01 | 69.1 | |
| DeepSeekMoE Chat 16B# Shot=5-shot, Total Params=16.4B, Activated Params=2.8B, FLOPs per 4K Tokens=74.4T2024.01 | 68.2 | |
| DeepSeek Chat 7B# Shot=5-shot, Total Params=6.9B, Activated Params=6.9B, FLOPs per 4K Tokens=183.5T2024.01 | 66.2 | |
| GShard 137B# Shot=5-shot2024.01 | 65.7 | |
| LLaMA2 7B# Shot=5-shot, # Total Params=6.7B, # Activated Params=6.7B, FLOPs per 4K Tokens=187.9T, # Training Tokens=2T2024.01 | 64 | |
| LLaMA2 SFT 7B# Shot=5-shot, Total Params=6.7B, Activated Params=6.7B, FLOPs per 4K Tokens=187.9T2024.01 | 48.4 |