General Knowledge Evaluation on MMLU
83.66MMLU AccuracyRADAR
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RADARAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 83.66 | — | — | — | — | |
| AgentPruneAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 82.4 | — | — | — | — | |
| MaASAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 82.32 | — | — | — | — | |
| AFlowAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 81.8 | — | — | — | — | |
| SC (COT×5)Autonomous Topology=false, Base LLM=gpt-4o-mini, Number of Agents=12026.05 | 80.66 | — | — | — | — | |
| LLM-DebateAutonomous Topology=false, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 80.56 | — | — | — | — | |
| G-DesignerAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 80.39 | — | — | — | — | |
| LLM-BlenderAutonomous Topology=false, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 80.29 | — | — | — | — | |
| DyLANAutonomous Topology=false, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 79.86 | — | — | — | — | |
| ComplexCoTAutonomous Topology=false, Base LLM=gpt-4o-mini, Number of Agents=12026.05 | 79.8 | — | — | — | — | |
| AutoAgentsAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 79.59 | — | — | — | — | |
| AgentSquareAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 79.58 | — | — | — | — | |
| MacNetAutonomous Topology=false, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 79.55 | — | — | — | — | |
| GTDAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 79.41 | — | — | — | — | |
| CoTAutonomous Topology=false, Base LLM=gpt-4o-mini, Number of Agents=12026.05 | 79.26 | — | — | — | — | |
| ARG-DesignerAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 79.1 | — | — | — | — | |
| DefaultModel=Qwen2.5-14B2026.02 | 78.9 | — | — | — | — | |
| CATModel=Qwen2.5-14B2026.02 | 78.7 | — | — | — | — | |
| Qwen2.5-14B-Instructvariant=baseline2026.03 | 78.69 | — | — | — | — | |
| VanillaAutonomous Topology=false, Base LLM=gpt-4o-mini, Number of Agents=12026.05 | 78.54 | — | — | — | — | |
| SIA-15B2026.03 | 78.45 | — | — | — | — | |
| AgentVerseAutonomous Topology=false, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 78.39 | — | — | — | — | |
| ADASAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 78.39 | — | — | — | — | |
| GPTSwarmAutonomous Topology=true, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 78.36 | — | — | — | — | |
| MixAT-GCGModel=Qwen2.5-14B2026.02 | 77.8 | — | — | — | — | |
| MultiPersonaAutonomous Topology=false, Base LLM=gpt-4o-mini, Number of Agents=52026.05 | 77.69 | — | — | — | — | |
| DATModel=Qwen2.5-14B2026.02 | 77.3 | — | — | — | — | |
| AttCTModel=Gemma-3-27B, Phase=Post2026.06 | 75.3 | — | — | — | — | |
| BCTModel=Gemma-3-27B, Phase=Post2026.06 | 74.2 | — | — | — | — | |
| Zero-shotBackbone=Qwen-3-8B2026.05 | 74.1 | — | — | — | — | |
| BCTModel=Qwen3-8B, Phase=Post2026.06 | 74.1 | — | — | — | — | |
| BCTModel=Qwen3-8B, Phase=Pre2026.06 | 74 | — | — | — | — | |
| ACTModel=Qwen3-8B, Phase=Pre2026.06 | 74 | — | — | — | — | |
| AttCTModel=Qwen3-8B, Phase=Pre2026.06 | 74 | — | — | — | — | |
| MLPCTModel=Qwen3-8B, Phase=Pre2026.06 | 74 | — | — | — | — | |
| BCTModel=Gemma-3-27B, Phase=Pre2026.06 | 73.8 | — | — | — | — | |
| ACTModel=Gemma-3-27B, Phase=Pre2026.06 | 73.8 | — | — | — | — | |
| ACTModel=Gemma-3-27B, Phase=Post2026.06 | 73.8 | — | — | — | — | |
| AttCTModel=Gemma-3-27B, Phase=Pre2026.06 | 73.8 | — | — | — | — | |
| MLPCTModel=Gemma-3-27B, Phase=Pre2026.06 | 73.8 | — | — | — | — | |
| MANSUBackbone=Qwen-3-8B, Unlearning Task=MUSE2026.05 | 73.7 | — | — | — | — | |
| ACTModel=Qwen3-8B, Phase=Post2026.06 | 73.7 | — | — | — | — | |
| AttCTModel=Qwen3-8B, Phase=Post2026.06 | 73.7 | — | — | — | — | |
| MLPCTModel=Qwen3-8B, Phase=Post2026.06 | 73.6 | — | — | — | — | |
| ReasonAnyBackbone=Qwen2.5-7B family2026.01 | 73.01 | — | — | — | — | |
| MANSUBackbone=Qwen-3-8B, Unlearning Task=WMDP-bio2026.05 | 72.9 | — | — | — | — | |
| MLPCTModel=Gemma-3-27B, Phase=Post2026.06 | 72.5 | — | — | — | — | |
| MANSUBackbone=Qwen-3-8B, Unlearning Task=WMDP-cyber2026.05 | 72.1 | — | — | — | — | |
| LEDBackbone=Qwen2.5-7B family2026.01 | 71.93 | — | — | — | — | |
| MANSUBackbone=Qwen-3-8B, Unlearning Task=WMDP-chem2026.05 | 71.4 | — | — | — | — | |
| BCTModel=Qwen3-4B, Phase=Post2026.06 | 68.6 | — | — | — | — | |
| BCTModel=Qwen3-4B, Phase=Pre2026.06 | 68.4 | — | — | — | — | |
| ACTModel=Qwen3-4B, Phase=Pre2026.06 | 68.4 | — | — | — | — | |
| AttCTModel=Qwen3-4B, Phase=Pre2026.06 | 68.4 | — | — | — | — | |
| MLPCTModel=Qwen3-4B, Phase=Pre2026.06 | 68.4 | — | — | — | — | |
| AttCTModel=Qwen3-4B, Phase=Post2026.06 | 68.2 | — | — | — | — | |
| AttCTModel=Llama-3.1-8B, Phase=Post2026.06 | 68 | — | — | — | — | |
| ACTModel=Qwen3-4B, Phase=Post2026.06 | 67.8 | — | — | — | — | |
| MLPCTModel=Llama-3.1-8B, Phase=Post2026.06 | 67.6 | — | — | — | — | |
| MLPCTModel=Qwen3-4B, Phase=Post2026.06 | 67.5 | — | — | — | — | |
| ACTModel=Llama-3.1-8B, Phase=Post2026.06 | 66.9 | — | — | — | — | |
| BCTModel=Llama-3.1-8B, Phase=Pre2026.06 | 66.4 | — | — | — | — | |
| BCTModel=Llama-3.1-8B, Phase=Post2026.06 | 66.4 | — | — | — | — | |
| ACTModel=Llama-3.1-8B, Phase=Pre2026.06 | 66.4 | — | — | — | — | |
| AttCTModel=Llama-3.1-8B, Phase=Pre2026.06 | 66.4 | — | — | — | — | |
| MLPCTModel=Llama-3.1-8B, Phase=Pre2026.06 | 66.4 | — | — | — | — | |
| wICIBackbone=Llama3.1-8B, Instruction Tuning Dataset=WizardLM, Training Budget=10% slice2026.04 | 64.9 | — | — | — | — | |
| Diffusion-onlyModel=Llama3-8B2026.02 | 64.4 | — | — | — | — | |
| DATModel=Llama3-8B2026.02 | 63.9 | — | — | — | — | |
| DefaultModel=Llama3-8B2026.02 | 63.8 | — | — | — | — | |
| CATModel=Llama3-8B2026.02 | 63.7 | — | — | — | — | |
| BaseAllowed Domains=—, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 63.6 | — | — | — | — | |
| CBModel=Llama3-8B2026.02 | 63.6 | — | — | — | — | |
| wICIBackbone=Llama3.1-8B, Instruction Tuning Dataset=Alpaca-GPT4, Training Budget=10% slice2026.04 | 63.45 | — | — | — | — | |
| PALETTEAllowed Domains=Sexual | Illegal, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 63.3 | — | — | — | — | |
| PALETTEAllowed Domains=Violence | Hate, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 62.7 | — | — | — | — | |
| PALETTEAllowed Domains=Hate | Disinfo, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 62.6 | — | — | — | — | |
| PALETTEAllowed Domains=Illegal | Violence, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 62.4 | — | — | — | — | |
| LATModel=Llama3-8B2026.02 | 61.9 | — | — | — | — | |
| PALETTEAllowed Domains=Disinfo | Sexual, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 61.8 | — | — | — | — | |
| GU+SimNPOFamily=Qwen, n=92026.05 | 61.7 | — | — | — | — | |
| MANSUFamily=Qwen, n=92026.05 | 61.7 | — | — | — | — | |
| MixAT-GCGModel=Llama3-8B2026.02 | 61.2 | — | — | — | — | |
| CASTAllowed Domains=Hate | Disinfo, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 61.1 | — | — | — | — | |
| CASTAllowed Domains=Violence | Hate, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 60.8 | — | — | — | — | |
| CASTAllowed Domains=Illegal | Violence, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 60.3 | — | — | — | — | |
| RMUModel Backbone=Llama-3-8B, Unlearning Method=RMU, RNA Configuration=Original2025.01 | 60.3 | — | — | — | — | |
| Zero-shotBackbone=Llama-3.1-8B-Instruct2026.05 | 60.3 | — | — | — | — | |
| RMU w/ RNAModel Backbone=Llama-3-8B, Unlearning Method=RMU, RNA Configuration=w/ RNA2025.01 | 60.2 | — | — | — | — | |
| CASTAllowed Domains=Sexual | Illegal, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 60.1 | — | — | — | — | |
| CASTAllowed Domains=Disinfo | Sexual, Backbone=LLAMA3-8B-INSTRUCT2026.05 | 59.4 | — | — | — | — | |
| NPOFamily=Qwen, n=92026.05 | 59.2 | — | — | — | — | |
| MANSUBackbone=Llama-3.1-8B-Instruct, Unlearning Task=MUSE2026.05 | 59.1 | — | — | — | — | |
| AttCTModel=Gemma-3-4B, Phase=Post2026.06 | 59 | — | — | — | — | |
| RMU w/ RNAModel Backbone=Mistral-7B, Unlearning Method=RMU, RNA Configuration=w/ RNA2025.01 | 58.6 | — | — | — | — | |
| MANSUBackbone=Llama-3.1-8B-Instruct, Unlearning Task=WMDP-cyber2026.05 | 58.6 | — | — | — | — | |
| ACTModel=Gemma-3-4B, Phase=Post2026.06 | 58.5 | — | — | — | — | |
| MANSUBackbone=Llama-3.1-8B-Instruct, Unlearning Task=WMDP-chem2026.05 | 58.4 | — | — | — | — | |
| BCTModel=Gemma-3-4B, Phase=Pre2026.06 | 58.4 | — | — | — | — | |
| ACTModel=Gemma-3-4B, Phase=Pre2026.06 | 58.4 | — | — | — | — |