Mathematical Reasoning on MATH-lighteval (M-LEval) (test)
95.6AccuracyAgent 1
Evaluation Results
| Method | Links | |
|---|---|---|
| Agent 1Controller Backbone=Qwen3-4B-Base, Agent 1 Model=Qwen3-30B-A3B-Instruct, Agent 2 Model=Qwen2.5-7B-Instruct, Agent 3 Model=Qwen2.5-1.5B-Instruct2026.05 | 95.6 | |
| Agent 1Controller Backbone=Qwen3-8B-Base, Agent 1 Model=Qwen3-30B-A3B-Instruct2026.05 | 95.6 | |
| Iterative Critique-and-Routing ControllerController Backbone=Qwen3-8B-Base2026.05 | 91.5 | |
| Iterative Critique-and-Routing ControllerController Backbone=Qwen3-4B-Base2026.05 | 90.5 | |
| Controller V2Controller Backbone=Qwen3-4B-Base2026.05 | 88.3 | |
| Router-R1Controller Backbone=Qwen3-4B-Base2026.05 | 81.1 | |
| Controller V2Controller Backbone=Qwen3-8B-Base2026.05 | 81 | |
| RouterDCController Backbone=Qwen3-4B-Base2026.05 | 78.4 | |
| RoBERTa RouterController Backbone=Qwen3-4B-Base2026.05 | 76.3 | |
| Controller V1Controller Backbone=Qwen3-4B-Base2026.05 | 76.1 | |
| Controller V1Controller Backbone=Qwen3-8B-Base2026.05 | 76 | |
| Agent 2Controller Backbone=Qwen3-4B-Base, Agent 2 Model=Qwen2.5-7B-Instruct2026.05 | 75.7 | |
| Random RouterController Backbone=Qwen3-4B-Base2026.05 | 75.6 | |
| Agent 2Controller Backbone=Qwen3-8B-Base, Agent 2 Model=Ministral-3-8B-Instruct2026.05 | 72.4 | |
| RouterDCController Backbone=Qwen3-8B-Base2026.05 | 68.1 | |
| Random RouterController Backbone=Qwen3-8B-Base2026.05 | 67 | |
| RoBERTa RouterController Backbone=Qwen3-8B-Base2026.05 | 65.9 | |
| Router-R1Controller Backbone=Qwen3-8B-Base2026.05 | 64 | |
| Agent 3Controller Backbone=Qwen3-4B-Base, Agent 3 Model=Qwen2.5-1.5B-Instruct2026.05 | 55.4 | |
| Agent 3Controller Backbone=Qwen3-8B-Base, Agent 3 Model=Llama-3-2-1B-Instruct2026.05 | 27.5 |