List your company

Question Answering on ARC-Challenge (Normalized Accuracy)

46.93Normalized AccuracyCALDERA
-2.917210.023922.96535.9061May 29, 2024Sep 7, 2024Dec 18, 2024Mar 30, 2025Jul 10, 2025Oct 20, 2025Jan 30, 2026

Evaluation results

MethodLinks
CALDERABackbone=Mistral 7B, Rank=256, Avg Bits=2.382024.0546.93
CALDERABackbone=Mistral 7B, Rank=128, Avg Bits=2.192024.0546.59
QuIP#Backbone=Mistral 7B, Rank=0, Avg Bits=22024.0544.8
SpanNormParam=A2.4B-16B, Tokens=200B, Architecture=MoE2026.0138.6
PreNormParam=A2.4B-16B, Tokens=200B, Architecture=MoE2026.0137
SpanNormParam=5B, Tokens=200B, Architecture=Dense2026.0136.3
PreNormParam=5B, Tokens=200B, Architecture=Dense2026.0134.4
SpanNormParam=1.3B, Tokens=100B, Architecture=Dense2026.0128.8
PreNormParam=1.3B, Tokens=100B, Architecture=Dense2026.0126.8
PreNormParam=740M, Tokens=30B, Architecture=Dense2026.0125.3
SpanNormParam=740M, Tokens=30B, Architecture=Dense2026.0125.1
AutoregressiveAlpha (α)=1, Data repetitions=12025.125.9
DualAlpha (α)=63/64, Data repetitions=12025.125.7
AutoregressiveAlpha (α)=1, Data repetitions=322025.125
DualAlpha (α)=3/4, Data repetitions=322025.123.3
DualAlpha (α)=1/8, Data repetitions=1282025.121.7
AutoregressiveAlpha (α)=1, Data repetitions=1282025.12-1