Reasoning on LiveBench
31.1AccuracyQwen2.5-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-7BModel Backbone=Qwen2.5-7B2026.06 | 31.1 | |
| PerSynStudent Model=Qwen2.5-3B2025.10 | 22.3 | |
| CARStudent Model=Qwen2.5-3B2025.10 | 20.9 | |
| Family-StrongStudent Model=Qwen2.5-3B2025.10 | 20.4 | |
| MixStudent Model=Qwen2.5-3B2025.10 | 19.3 | |
| StrongStudent Model=Qwen2.5-3B2025.10 | 19.1 | |
| PerSynStudent Model=Qwen2.5-1.5B2025.10 | 14.8 | |
| Family-StrongStudent Model=Qwen2.5-1.5B2025.10 | 13.6 | |
| MixStudent Model=Qwen2.5-1.5B2025.10 | 13.3 | |
| CARStudent Model=Qwen2.5-1.5B2025.10 | 13.3 | |
| StrongStudent Model=Qwen2.5-1.5B2025.10 | 12.8 | |
| CARStudent Model=Gemma-2-2B2025.10 | 12.8 | |
| PerSynStudent Model=Llama-3.2-3B2025.10 | 12.6 | |
| PerSynStudent Model=Gemma-2-2B2025.10 | 12.4 | |
| Qwen2.5-0.5BModel Backbone=Qwen2.5-0.5B2026.06 | 12.1 | |
| MixStudent Model=Llama-3.2-3B2025.10 | 12 | |
| CARStudent Model=Llama-3.2-3B2025.10 | 11.8 | |
| Family-StrongStudent Model=Gemma-2-2B2025.10 | 11.6 | |
| Family-StrongStudent Model=Llama-3.2-3B2025.10 | 11.1 | |
| MixStudent Model=Gemma-2-2B2025.10 | 10.9 | |
| Uni-EnergyModel Backbone=LLaDa-8B, Decoding Strategy=Unified energy-based decoding2026.06 | 10.9 | |
| StrongStudent Model=Llama-3.2-3B2025.10 | 10.8 | |
| StrongStudent Model=Gemma-2-2B2025.10 | 10.3 | |
| Uni-EnergyModel Backbone=Dream-7B, Decoding Strategy=Unified energy-based decoding2026.06 | 9.9 | |
| PerSynStudent Model=Qwen2.5-0.5B2025.10 | 9.8 | |
| Ind-EnergyModel Backbone=Dream-7B, Decoding Strategy=Independent energy-based decoding2026.06 | 9.3 | |
| CARStudent Model=Qwen2.5-0.5B2025.10 | 9 | |
| Inv-EnergyModel Backbone=Dream-7B, Decoding Strategy=Invariant energy-based decoding2026.06 | 8.8 | |
| Family-StrongStudent Model=Qwen2.5-0.5B2025.10 | 8.6 | |
| Ind-EnergyModel Backbone=LLaDa-8B, Decoding Strategy=Independent energy-based decoding2026.06 | 8.6 | |
| StrongStudent Model=Qwen2.5-0.5B2025.10 | 8.4 | |
| APDModel Backbone=Dream-7B, Decoding Strategy=Speculative decoding2026.06 | 8.2 | |
| MixStudent Model=Qwen2.5-0.5B2025.10 | 8.1 | |
| Dream-7BModel Backbone=Dream-7B, Decoding Strategy=Base2026.06 | 7.1 | |
| COREModel Backbone=Dream-7B, Decoding Strategy=Remask decoding2026.06 | 6.5 | |
| APDModel Backbone=LLaDa-8B, Decoding Strategy=Speculative decoding2026.06 | 6.3 | |
| Inv-EnergyModel Backbone=LLaDa-8B, Decoding Strategy=Invariant energy-based decoding2026.06 | 5.7 | |
| DAWNModel Backbone=LLaDa-8B, Decoding Strategy=Dependency decoding2026.06 | 4.8 | |
| LLaDa-8BModel Backbone=LLaDa-8B, Decoding Strategy=Base2026.06 | 4.1 | |
| COREModel Backbone=LLaDa-8B, Decoding Strategy=Remask decoding2026.06 | 3 |