List your company

Prompt-heavy Aggregate Performance Evaluation on Qasper, Multi-News, HotpotQA, Musique, and 2WikiMQA

64.81Weighted Top-1CASK
57.041259.058161.07563.0919Apr 13, 2026

Evaluation results

MethodLinks
CASKBudget=3842026.0464.8189.41.521
CASKBudget=2562026.0463.7288.591.594
TriAttentionBudget=2562026.0458.4282.741.985
TriAttentionBudget=3842026.0457.3482.742.002