List your company
ResearchFormerly WizWand

Common Sense Reasoning on PIQA (dev)

83.2AccuracyMegatron-NLG
71.749674.722377.69580.6677Dec 13, 2021Aug 31, 2022May 20, 2023Feb 5, 2024Oct 24, 2024Jul 12, 2025Mar 31, 2026
Updated 7mo ago

Evaluation results

MethodLinks
Megatron-NLGModel Size=530B, Evaluation Protocol=Few-shot2021.1283.2
GPT-3Model Size=175B, Evaluation Protocol=Few-shot, Shots=502021.1282.3
GopherModel Size=280B, Evaluation Protocol=Few-shot, Shots=02021.1281.8
GLaMModel Size=64B/64E, Evaluation Protocol=Few-shot, Shots=322021.1281.8
GLaMModel Size=64B/64E, Evaluation Protocol=One-shot2021.1281.4
GPT-3Model Size=175B, Evaluation Protocol=Zero-shot2021.1281
GPT-3Model Size=175B, Evaluation Protocol=One-shot2021.1280.5
GLaMModel Size=64B/64E, Evaluation Protocol=Zero-shot2021.1280.4
CKTModel Scale=large2023.0676.07
CALMModel Scale=large2023.0675.11
Content-SharpRouterDiscovery Source=AI Discovered2026.0374.37
Hier-GateNetDiscovery Source=AI Discovered2026.0374.37
Gated DeltaNetDiscovery Source=Human Discovered2026.0374.1
AdaMulti-PathGateNetDiscovery Source=AI Discovered2026.0374.1
Mamba2Discovery Source=Human Discovered2026.0373.78
DeltaNetDiscovery Source=Human Discovered2026.0373.12
PathGate-FusionNetDiscovery Source=AI Discovered2026.0372.91
FusionGated-FIRNetDiscovery Source=AI Discovered2026.0372.91
T5Model Scale=large2023.0672.19