List your company
ResearchFormerly WizWand

Question Answering on ARC

94.6AccuracyYi-34B + RTD
62.682470.968779.25587.5413May 31, 2024Sep 16, 2024Jan 2, 2025Apr 20, 2025Aug 6, 2025Nov 22, 2025Mar 10, 2026
Updated 7mo ago

Evaluation results

MethodLinks
Yi-34B + RTDevaluation=5-shot2024.0994.6
Yi-34Bevaluation=5-shot ICL2024.0994
Yi-34Bevaluation=zero-shot2024.0993.3
Dense ModelBase Model=Llama3.1-8B, Fine-tuning Protocol=LoRA, Sparsity=0%2026.0193.1
GradPrunerBase Model=Llama3.1-8B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0192.3
LLaMA2-70B + RTDevaluation=5-shot2024.0991.7
LLaMA2-70Bevaluation=5-shot ICL2024.0991.6
Dense ModelBase Model=Mistral-7B, Fine-tuning Protocol=LoRA, Sparsity=0%2026.0190.9
Yi-34B + RTDevaluation=zero-shot2024.0990.7
Dense ModelBase Model=Mistral-7B, Fine-tuning Protocol=FFT, Sparsity=0%2026.0190.5
FT(Llama3.2)Base Model=Llama3.2-3B, Fine-tuning Protocol=LoRA, Sparsity=0%2026.0190.2
GradPrunerBase Model=Mistral-7B, Fine-tuning Protocol=FFT, Sparsity=40%2026.0189.5
GradPrunerBase Model=Mistral-7B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0189.3
APTBase Model=Llama3.1-8B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0189.1
Sonnet-3.5Retrieval=None2026.0388.5
SATBase Model=Llama3.1-8B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0188.2
LacoBase Model=Llama3.1-8B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0187.5
FT(Llama3.2)Base Model=Llama3.2-3B, Fine-tuning Protocol=FFT, Sparsity=0%2026.0187.4
LLaMA3-8Bevaluation=5-shot ICL2024.0987.3
MINITRONBase Model=Llama3.1-8B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0187.3
LLaMA3-8B + RTDevaluation=5-shot2024.0987.1
LLMPrunerBase Model=Llama3.1-8B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0187
LLMPrunerBase Model=Mistral-7B, Fine-tuning Protocol=FFT, Sparsity=40%2026.0186.7
SATBase Model=Mistral-7B, Fine-tuning Protocol=FFT, Sparsity=40%2026.0186.7
APTBase Model=Mistral-7B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0186.6
Dense ModelBase Model=Llama3.1-8B, Fine-tuning Protocol=FFT, Sparsity=0%2026.0186.5
LacoBase Model=Mistral-7B, Fine-tuning Protocol=FFT, Sparsity=40%2026.0186.4
Qwen3-32B+Hit-RAGFramework=Hit-RAG2026.0386.2
LLaMA2-70B + RTDevaluation=zero-shot2024.0986.1
SATBase Model=Mistral-7B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0185.7
MINITRONBase Model=Mistral-7B, Fine-tuning Protocol=FFT, Sparsity=40%2026.0185.1
LacoBase Model=Mistral-7B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0185.1
Llama3.1-70B+R-InsRetrieval=Yes2026.0385.1
GradPrunerBase Model=Llama3.1-8B, Fine-tuning Protocol=FFT, Sparsity=40%2026.0184.9
Llama3.1-70B+Hit-RAGFramework=Hit-RAG2026.0384.8
LLMPrunerBase Model=Mistral-7B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0184
MINITRONBase Model=Mistral-7B, Fine-tuning Protocol=LoRA, Sparsity=40%2026.0184
LLaMA3-8B + RTDevaluation=zero-shot2024.0983.7
MINITRONBase Model=Llama3.1-8B, Fine-tuning Protocol=FFT, Sparsity=40%2026.0182.6
LacoBase Model=Llama3.1-8B, Fine-tuning Protocol=FFT, Sparsity=40%2026.0182.4
SATBase Model=Llama3.1-8B, Fine-tuning Protocol=FFT, Sparsity=40%2026.0182.2
MetaICLk=40, Backbone=Llama-2-7b-chat-hf2025.0182
MetaICLk=MAX, Backbone=Llama-2-7b-chat-hf2025.0182
LLMPrunerBase Model=Llama3.1-8B, Fine-tuning Protocol=FFT, Sparsity=40%2026.0181.7
Qwen3-4B+Hit-RAGFramework=Hit-RAG2026.0381.4
Qwen3-8B+Hit-RAGFramework=Hit-RAG2026.0381.3
DrICLk=5, Backbone=Llama-2-7b-chat-hf2025.0181
DrICLk=30, Backbone=Llama-2-7b-chat-hf2025.0181
DrICLk=MAX, Backbone=Llama-2-7b-chat-hf2025.0181
Llama3.1-8B+Hit-RAGFramework=Hit-RAG2026.0380.2
Llama3.1-8B+R-InsRetrieval=Yes2026.0379.9
DrICLk=AVG, Backbone=Llama-2-7b-chat-hf2025.0178
GLM3-6B + RTDevaluation=5-shot2024.0976.5
Speculative RAG-7BRetrieval=Yes2026.0376.2
MetaICLk=AVG, Backbone=Llama-2-7b-chat-hf2025.0176
GLM3-6Bevaluation=5-shot ICL2024.0975.3
ChatGPTRetrieval=None2026.0375.3
Ret-ChatGPTRetrieval=None2026.0375.3
GLM3-6B + RTDevaluation=zero-shot2024.0975
Self-RAG-13BRetrieval=Yes2026.0373.1
RAS-8BRetrieval=Yes2026.0371.4
LLaMA3-8Bevaluation=zero-shot2024.0971.2
ITk=0, Backbone=Llama-2-7b-chat-hf2025.0171
ITk=MAX, Backbone=Llama-2-7b-chat-hf2025.0171
BaseTraining Stage=SFT-1B, Evaluation Protocol=Zero-shot2025.1269.65
Qwen3-32BRetrieval=None2026.0369.4
SOLARTraining Stage=SFT-1B, Evaluation Protocol=Zero-shot2025.1269.28
MSSRfullBackbone=Llama-3.1-8B2026.0369.2
SOLARTraining Stage=CPT-1B, Evaluation Protocol=Zero-shot2025.1268.81
MSSRschBackbone=Llama-3.1-8B2026.0368.8
BaseTraining Stage=CPT-1B, Evaluation Protocol=Zero-shot2025.1268.69
RAS-7BRetrieval=Yes2026.0368.5
Llama ProTraining Stage=SFT-1B, Evaluation Protocol=Zero-shot2025.1267.68
LLaMA2-70Bevaluation=zero-shot2024.0967.4
LLaMA2-7B + RTDevaluation=5-shot2024.0967.3
Self-RAG-7BRetrieval=Yes2026.0367.3
Avg-DeUSTraining Stage=SFT-1B, Evaluation Protocol=Zero-shot2025.1267.09
Avg-DeUSTraining Stage=CPT-1B, Evaluation Protocol=Zero-shot2025.1266.84
UniBiasModel=Llama-2 13b, ICL Setting=1-shot2024.0566.81
Llama ProTraining Stage=CPT-1B, Evaluation Protocol=Zero-shot2025.1266.75
OpT-DeUSTraining Stage=SFT-1B, Evaluation Protocol=Zero-shot2025.1266.62
ICLModel=Llama-2 13b, ICL Setting=1-shot2024.0566.54
OpT-DeUSTraining Stage=CPT-1B, Evaluation Protocol=Zero-shot2025.1266.46
LESATraining Stage=CPT-1B, Evaluation Protocol=Zero-shot2025.1266.33
MIDUS-HMLTraining Stage=SFT-1B, Evaluation Protocol=Zero-shot2025.1266.33
MIDUS-HMLTraining Stage=CPT-1B, Evaluation Protocol=Zero-shot2025.1266.16
LESATraining Stage=SFT-1B, Evaluation Protocol=Zero-shot2025.1265.99
LLaMA2-7Bevaluation=5-shot ICL2024.0965
NFTk=0, Backbone=Llama-2-7b-chat-hf2025.0165
NFTk=MAX, Backbone=Llama-2-7b-chat-hf2025.0165
MSSRsplBackbone=Llama-3.1-8B2026.0365
DCModel=Llama-2 13b, ICL Setting=1-shot2024.0564.88
SPACEIteration=Iter22025.1264.85
SPACEIteration=Iter42025.1264.85
SPACEIteration=Iter32025.1264.59
AccuBackbone=Llama-3.1-8B2026.0364.5
CCModel=Llama-2 13b, ICL Setting=1-shot2024.0564.33
SPINIteration=Iter12025.1264.33
SPINIteration=Iter42025.1263.99
S-IPOIteration=Iter32025.1263.91