List your company

Reasoning on MMLU-Pro (Subject Averages)

55Average ScoreRAT
22.03230.59139.1547.709Aug 8, 2025Oct 1, 2025Nov 25, 2025Jan 19, 2026Mar 14, 2026May 8, 2026Jul 2, 2026

Evaluation results

MethodLinks
RATBackbone=GPT-4o-mini, Category=Advanced RAG Methods2025.085557.555.37334.2
CoTBackbone=GPT-4o-mini2025.0854.556.753.170.438.2
Standard RAGBackbone=GPT-4o-mini2025.0853.55752.368.636.2
Search-o1Backbone=GPT-4o-mini, Category=Advanced RAG Methods2025.0853.553.555.369.835.4
UR2 (GPT Summary)Backbone=Qwen-2.5-7B2025.0853.353.25372.235
Self-AskBackbone=GPT-4o-mini, Category=Advanced RAG Methods2025.085156.348.567.831.2
UR2 (qw Summary)Backbone=Qwen-2.5-7B2025.0850.351.249.767.832.6
Open-Reasoner-ZeroBackbone=Qwen-2.5-7B, Category=CoT-RL Methods2025.0849.65046.667.534.2
Vanilla RLBackbone=Qwen-2.5-7B2025.0848.452.243.56433.8
UR2 (GPT summary)Backbone=Qwen-2.5-3B2025.0847.847.849.363.930
UR2Model=Qwen-2.5-3B2025.0847.847.849.363.930
General ReasonerBackbone=Qwen-2.5-7B, Category=CoT-RL Methods2025.0847.147.944.265.930.4
RATModel=Qwen-2.5-7B2025.0846.647.244.764.430
Search-o1Model=Qwen-2.5-7B2025.0845.442.845.963.229.6
NeuFSModel=Qwen3-4B, InfoType=Neuron, Aggregated Shots=5, 10, 20, 302026.0745.2————
CoTBackbone=Qwen-2.5-7B2025.0844.542.345.763.426.6
TypiClust (S.)Model=Qwen3-4B, InfoType=S., Aggregated Shots=5, 10, 20, 302026.0743.7————
UR2 (GPT summary)Backbone=LLaMA-3.1-8B2025.0843.448.338.65828.8
UR2Model=LLaMA-3.1-8B2025.0843.448.338.65828.8
UR2 (qw summary)Backbone=Qwen-2.5-3B2025.0843.24342.660.626.6
Entropy (Highest)Model=Qwen3-4B, InfoType=Highest, Aggregated Shots=5, 10, 20, 302026.0743————
Standard RAGBackbone=Qwen-2.5-7B2025.0842.444.641.157.826
UR2 (qw summary)Backbone=LLaMA-3.1-8B2025.0842.443.641.256.828
Self-AskModel=Qwen-2.5-7B2025.0842.340.742.16026.2
FastVoteK (S.+L.)Model=Qwen3-4B, InfoType=S.+L., Aggregated Shots=5, 10, 20, 302026.0742————
Entropy (Diverse)Model=Qwen3-4B, InfoType=Diverse, Aggregated Shots=5, 10, 20, 302026.0741.4————
RandomModel=Qwen3-4B, InfoType=N/A, Aggregated Shots=5, 10, 20, 302026.0741.2————
CoTBackbone=LLaMA-3.1-8B2025.0840.337.840.953.429
CoTModel=LLaMA-3.1-8B2025.0840.337.840.953.429
RATModel=LLaMA-3.1-8B2025.0840.342.337.752.628.6
Vanilla RLBackbone=LLaMA-3.1-8B2025.0840.244.636.95326.4
Vanilla RLModel=LLaMA-3.1-8B2025.0840.244.636.95326.4
FastVoteK (L.)Model=Qwen3-4B, InfoType=L., Aggregated Shots=5, 10, 20, 302026.0740.1————
VoteK (S.+E.)Model=Qwen3-4B, InfoType=S.+E., Aggregated Shots=5, 10, 20, 302026.0740.1————
TypiClust (S.+L.)Model=Qwen3-4B, InfoType=S.+L., Aggregated Shots=5, 10, 20, 302026.0739.8————
PatronModel=Qwen3-4B, InfoType=S.+E., Aggregated Shots=5, 10, 20, 302026.0739.1————
Standard RAGBackbone=LLaMA-3.1-8B2025.0838.843.633.95126.6
Standard RAGModel=LLaMA-3.1-8B2025.0838.843.633.95126.6
Vanilla RLBackbone=Qwen-2.5-3B2025.0838.740.734.75524.6
Vanilla RLModel=Qwen-2.5-3B2025.0838.740.734.75524.6
FastVoteK (S.)Model=Qwen3-4B, InfoType=S., Aggregated Shots=5, 10, 20, 302026.0738.6————
SimpleRL-ZooBackbone=Qwen-2.5-7B, Category=CoT-RL Methods2025.0838.335.736.955.225.4
TypiClust (L.)Model=Qwen3-4B, InfoType=L., Aggregated Shots=5, 10, 20, 302026.0737.6————
VoteK (S.+E.+L.)Model=Qwen3-4B, InfoType=S.+E.+L., Aggregated Shots=5, 10, 20, 302026.0737.4————
Standard RAGBackbone=Qwen-2.5-3B2025.0837.137.836.551.423.2
Standard RAGModel=Qwen-2.5-3B2025.0837.137.836.551.423.2
Self-AskModel=LLaMA-3.1-8B2025.0835.639.832.14723.4
RATModel=Qwen-2.5-3B2025.0835.137.833.550.618.4
Search-o1Model=Qwen-2.5-3B2025.0834.933.934.550.620.6
Search-o1Model=LLaMA-3.1-8B2025.0834.832.632.54628
CoTBackbone=Qwen-2.5-3B2025.0833.833.632.348.820.6
CoTModel=Qwen-2.5-3B2025.0833.833.632.348.820.6
Self-AskModel=Qwen-2.5-3B2025.083233.130.544.220.2
NeuFSModel=Llama-3.2-3B, InfoType=Neuron, Aggregated Shots=5, 10, 20, 302026.0725.1————
RandomModel=Llama-3.2-3B, InfoType=N/A, Aggregated Shots=5, 10, 20, 302026.0724.9————
TypiClust (L.)Model=Llama-3.2-3B, InfoType=L., Aggregated Shots=5, 10, 20, 302026.0724.8————
VoteK (S.+E.)Model=Llama-3.2-3B, InfoType=S.+E., Aggregated Shots=5, 10, 20, 302026.0724.6————
Entropy (Highest)Model=Llama-3.2-3B, InfoType=Highest, Aggregated Shots=5, 10, 20, 302026.0724.5————
FastVoteK (S.)Model=Llama-3.2-3B, InfoType=S., Aggregated Shots=5, 10, 20, 302026.0724.5————
PatronModel=Llama-3.2-3B, InfoType=S.+E., Aggregated Shots=5, 10, 20, 302026.0724.4————
Entropy (Diverse)Model=Llama-3.2-3B, InfoType=Diverse, Aggregated Shots=5, 10, 20, 302026.0724.4————
TypiClust (S.+L.)Model=Llama-3.2-3B, InfoType=S.+L., Aggregated Shots=5, 10, 20, 302026.0724.3————
TypiClust (S.)Model=Llama-3.2-3B, InfoType=S., Aggregated Shots=5, 10, 20, 302026.0724.2————
FastVoteK (L.)Model=Llama-3.2-3B, InfoType=L., Aggregated Shots=5, 10, 20, 302026.0724.2————
FastVoteK (S.+L.)Model=Llama-3.2-3B, InfoType=S.+L., Aggregated Shots=5, 10, 20, 302026.0724.2————
VoteK (S.+E.+L.)Model=Llama-3.2-3B, InfoType=S.+E.+L., Aggregated Shots=5, 10, 20, 302026.0723.3————