Share your thoughts, 1 month free Claude Pro on usSee more

Reading Comprehension on SciQ

93.7Accuracy

LLaMA2-7B

Updated 3mo ago

Evaluation Results

Method	Links
LLaMA2-7B 2023.10		93.7
Open-LLaMA-3B-v2 2023.10		91.8
Open-LLaMA-3B-v1 2023.10		91.3
Sheared-LLaMA-2.7B 2023.10		90.8
INCITE-Base-3B 2023.10		90.7
TinyLlama-1.1B 2023.10		88.9
Pythia-2.8B 2023.10		88.3
Sheared-LLaMA-1.3B 2023.10		87.3
Pythia-1.4B 2023.10		86.4
OPT-2.7B 2023.10		85.8
OPT-1.3B 2023.10		84.3
DoGraph 2026.04		66.1
RegMix 2026.04		62.8
DoReMi 2026.04		61.6
Uniform 2026.04		60.3
DOGE 2026.04		60.1
Dynamic Loss-Based 2026.04		60
Data Mixing Law 2026.04		58.6
Dynamic Loss-Based 2026.04		53.8
DoGraph 2026.04		53.8
DoGraph 2026.04		53.5
Dynamic Loss-Based 2026.04		53.3
DoReMi 2026.04		53.3
DOGE 2026.04		52.8
DoReMi 2026.04		52.4
Uniform 2026.04		49
RegMix 2026.04		46.6
Data Mixing Law 2026.04		45.2
Data Mixing Law 2026.04		39.6
RegMix 2026.04		38.7
Uniform 2026.04		32.6
DOGE 2026.04		31.1