List your company
ResearchFormerly WizWand

Question Answering on ARC-E (Accuracy, Token, Speed)

91.2AccuracyLatentMAS
78.7281.9685.288.44Nov 25, 2025
Updated 5mo ago

Evaluation results

MethodLinks
LatentMASBackbone=Llama-3.1-8B-Instruct, MAS Setting=Sequential2025.1191.2228516
LatentMASBackbone=Llama-3.1-8B-Instruct, MAS Setting=Hierarchical2025.1190.4206669
TextMASBackbone=Llama-3.1-8B-Instruct, MAS Setting=Sequential2025.1189.61,9812,708
TextMASBackbone=Llama-3.1-8B-Instruct, MAS Setting=Hierarchical2025.1189.31,1181,832
SingleBackbone=Llama-3.1-8B-Instruct, MAS Setting=Sequential2025.1185.7219352
SingleBackbone=Llama-3.1-8B-Instruct, MAS Setting=Hierarchical2025.1185.7219352
LatentMASBackbone=Llama-3.2-3B-Instruct, MAS Setting=Hierarchical2025.1182.3177192
TextMASBackbone=Llama-3.2-3B-Instruct, MAS Setting=Hierarchical2025.1181.6956615
LatentMASBackbone=Llama-3.2-3B-Instruct, MAS Setting=Sequential2025.1180.4229392
TextMASBackbone=Llama-3.2-3B-Instruct, MAS Setting=Sequential2025.1180.11,8891,670
SingleBackbone=Llama-3.2-3B-Instruct, MAS Setting=Sequential2025.1179.2236155
SingleBackbone=Llama-3.2-3B-Instruct, MAS Setting=Hierarchical2025.1179.2236155