Commonsense Reasoning on PIQA (test)

90.1Accuracy

UNICORN

Updated 22d ago

Evaluation Results

Method	Links
UNICORN 2022.01		90.1
MT-NLG 2022.01		83.19
GPT-3 2022.01		82.3
MT-NLG 2022.01		81.99
Gopher 2022.01		81.8
Qwen3-4B + WeMask(SFT) 2026.05		81.23
Dense 2023.01		81.07
Qwen3-4B + SFT + WeMask(TF) 2026.05		81.03
Qwen3-4B + SFT + WeMask(TF) 2026.05		81.01
Qwen3-4B + WeMask(SFT) 2026.05		81.01
GPT-3 2022.01		81
MT-NLG 2022.01		80.96
Qwen3-4B + WeMask(SFT) 2026.05		80.96
Qwen3-4B + SFT + WeMask(TF) 2026.05		80.69
SparseGPT 2023.01		80.63
Qwen3-4B + SFT + WeMask(TF) 2026.05		80.52
GPT-3 2022.01		80.5
Qwen3-4B + WeMask(SFT) 2026.05		80.47
Qwen3-4B + SFT 2026.05		80.3
SparseGPT 2023.01		79.54
SparseGPT 2023.01		79.54
GPT-3 175B 2021.10		79.27
Coherence Boosting (GPT-3 175B) 2021.10		78.94
Qwen3-4B + WeMask(SFT) 2026.05		77.2
Pythia-12B 2023.04		76
Pythia-6.9B 2023.04		75.2
Pythia-2.8B 2023.04		73.9
Dip-SVD 2026.02		73.4
SAES-SVD 2026.02		73.4
SVD-LLM 2026.02		72.5
Coherence Boosting (GPT-2 XL) 2021.10		71.49
Pythia-1.4B 2023.04		71.1
Muon+CWD 2025.10		71
GPT-2 XL (1.6B) 2021.10		70.84
Pythia-1B 2023.04		70.7
AdamW+CWD 2025.10		69
Muon 2025.10		68
AdamW 2025.10		67
Pythia-410M 2023.04		66.8
GPT-3 175B 2021.10		66.32
Dip-SVD 2026.02		66.1
SVD-LLM 2026.02		63.7
Coherence Boosting (GPT-2 Small) 2021.10		63.44
SAES-SVD 2026.02		63.1
GPT-2 Small (125M) 2021.10		62.89
Pythia-160M 2023.04		62.7
GPT-2 XL (1.6B) 2021.10		60.45
PGM 6 / 6 (1024) 2025.05		59.85
Pythia-70M 2023.04		59.5
PGM 6 / 6 (1024) 2025.05		59.19
PGM 8 / 8 2025.05		59.03
PGM 8 / 8 2025.05		58.92
MDLM 2025.05		58.27
MDLM 2025.05		57.73
GPT-2 Small (125M) 2021.10		57.45
Magnitude 2023.01		54.73
Qwen3-4B + SFT + WeMask(TF) 2026.05		49.44