List your company
ResearchFormerly WizWand

General Knowledge Question Answering on MMLU

82.06AccuracyIoT
19.285635.582851.8868.1772Mar 28, 2025Jun 6, 2025Aug 15, 2025Oct 24, 2025Jan 2, 2026Mar 13, 2026May 23, 2026
Updated 5mo ago

Evaluation results

MethodLinks
IoTModel=GPT-4o mini2026.0382.06
CoTModel=GPT-4o mini2026.0381.31
QWEN-2.5-7BBackbone=QWEN-2.5-7B2026.0579.7
Ellipsoid ControlBackbone=QWEN-2.5-7B2026.0579.4
LoRA-SteerBackbone=QWEN-2.5-7B2026.0578.8
PGD-SphereBackbone=QWEN-2.5-7B2026.0578.4
HSFBackbone=QWEN-2.5-7B2026.0578
Jailbreak AntidoteBackbone=QWEN-2.5-7B2026.0576.5
IoTModel=Olmo-2-13B2026.0371.34
CoTModel=Olmo-2-13B2026.0370.59
SCModel=Olmo-2-13B2026.0369.86
EoTModel=Olmo-2-13B2026.0369.62
SCModel=Llama-3.3-8B2026.0369
LLAMA3-8BBackbone=LLAMA3-8B2026.0568.5
Circuit BreakerBackbone=LLAMA3-8B2026.0568.3
Ellipsoid ControlBackbone=LLAMA3-8B2026.0568.3
HSFBackbone=LLAMA3-8B2026.0568.2
Jailbreak AntidoteBackbone=LLAMA3-8B2026.0567.8
IoTModel=Olmo-2-7B2026.0366.78
EoTModel=Llama-3.3-8B2026.0366.56
IoTModel=Llama-3.3-8B2026.0366.31
CoTModel=Olmo-2-7B2026.0365.62
Mistral-7B-v2Backbone=Mistral-7B-v22026.0565.6
Circuit BreakerBackbone=Mistral-7B-v22026.0565.6
CoTModel=Llama-3.3-8B2026.0365.56
Ellipsoid ControlBackbone=Mistral-7B-v22026.0565.5
LoRA-SteerBackbone=Mistral-7B-v22026.0565.4
HSFBackbone=Mistral-7B-v22026.0565
SCModel=Olmo-2-7B2026.0364.91
PGD-SphereBackbone=Mistral-7B-v22026.0564.5
Jailbreak AntidoteBackbone=Mistral-7B-v22026.0564.3
EoTModel=Olmo-2-7B2026.0362.62
ME-DLM Stage 3Budget=1/1, Evaluation Protocol=fully generative2026.0562.6
LoTVerifier Training Dataset=AQuA2025.0362.3
ME-DLM Stage 3Budget=1/2, Evaluation Protocol=fully generative2026.0561.5
ME-DLM Stage 2Budget=1/1, Evaluation Protocol=fully generative2026.0560.6
LLaDABudget=1/1, Evaluation Protocol=fully generative2026.0560.2
ME-DLM Stage 3Budget=1/4, Evaluation Protocol=fully generative2026.0558.5
LoRA-SteerBackbone=LLAMA3-8B2026.0558.3
ME-DLM Stage 2Budget=1/2, Evaluation Protocol=fully generative2026.0557.7
LLaDABudget=1/2, Evaluation Protocol=fully generative2026.0557.1
ME-DLM Stage 3Budget=1/8, Evaluation Protocol=fully generative2026.0556.2
LoTVerifier Training Dataset=MMLU2025.0353
LoTVerifier Training Dataset=CommonSenseQA2025.0353
LLaDABudget=1/4, Evaluation Protocol=fully generative2026.0544.5
ME-DLM Stage 2Budget=1/4, Evaluation Protocol=fully generative2026.0543.3
LoTVerifier Training Dataset=StrategyQA2025.0340.5
PGD-SphereBackbone=LLAMA3-8B2026.0527.4
LLaDABudget=1/8, Evaluation Protocol=fully generative2026.0523.4
ME-DLM Stage 2Budget=1/8, Evaluation Protocol=fully generative2026.0521.7