List your company

Commonsense Reasoning on Hellaswag (HS and Accuracy)

50HS ScorePre-unlearn
-1.511211.861925.23538.6081Oct 1, 2025Oct 28, 2025Nov 24, 2025Dec 21, 2025Jan 17, 2026Feb 13, 2026Mar 12, 2026

Evaluation results

MethodLinks
Pre-unlearn2025.1050—
NPO w/ ZOAlgorithm=NPO, Optimizer=ZO2025.1044.7—
NoneModel=Mistral, Mixing ratio (r)=0.12026.0344.594.1
BeavertailsModel=Qwen2.5, Mixing ratio (r)=0.12026.0344.2292.55
BeavertailsModel=Mistral, Mixing ratio (r)=0.12026.0342.9193.83
GradDiff w/ signAdamAlgorithm=GradDiff, Optimizer=signAdam2025.1042.7—
GradDiff w/ signSGDAlgorithm=GradDiff, Optimizer=signSGD2025.1042.2—
GradDiff w/ SAMAlgorithm=GradDiff, Optimizer=SAM2025.1042.1—
GradDiffAlgorithm=GradDiff2025.1041.7—
GradDiff w/ RSAlgorithm=GradDiff, Optimizer=RS2025.1039.4—
GradDiff w/ ZOAlgorithm=GradDiff, Optimizer=ZO2025.1036.7—
NPO w/ signSGDAlgorithm=NPO, Optimizer=signSGD2025.1035.8—
NPO w/ signAdamAlgorithm=NPO, Optimizer=signAdam2025.1033.2—
NPO w/ RSAlgorithm=NPO, Optimizer=RS2025.1031.1—
NPOAlgorithm=NPO2025.1031—
GradDiff w/ HybridAlgorithm=GradDiff, Optimizer=Hybrid2025.1030.8—
PLDRv51-SOC-110M-5Zero-shot=true, Criticality=SOC-110M-52026.0330.55—
GPT-Neo-125MZero-shot=true2026.0330.4—
BeavertailsModel=Llama3, Mixing ratio (r)=0.12026.0330.3591.98
NPO w/ SAMAlgorithm=NPO, Optimizer=SAM2025.1029.7—
PLDRv51-SOC-110M-2Zero-shot=true, Criticality=SOC-110M-22026.0329.27—
PLDRv51-SOC-110M-4Zero-shot=true, Criticality=SOC-110M-42026.0329.13—
PLDRv51-SOC-110M-3Zero-shot=true, Criticality=SOC-110M-32026.0328.99—
PLDRv51-SOC-110M-1Zero-shot=true, Criticality=SOC-110M-12026.0328.65—
NPO w/ HybridAlgorithm=NPO, Optimizer=Hybrid2025.1028.4—
ABL-SOC-110M-1Zero-shot=true2026.0326.88—
SUB-SOC-110M-1Zero-shot=true2026.0326.03—
ABL-SOC-110M-2Zero-shot=true2026.0326.03—
SUB-SOC-110M-2Zero-shot=true2026.0325.93—
ABL-SOC-110M-3Zero-shot=true2026.0325.45—
BeavertailsModel=OLMo2, Mixing ratio (r)=0.12026.0316.4388.47
NoneModel=Qwen2.5, Mixing ratio (r)=0.12026.0315.1592.81
GR-SAPModel=Mistral, Mixing ratio (r)=0.12026.0314.1193.78
AegisModel=Mistral, Mixing ratio (r)=0.12026.0313.8293.7
GR-SAPModel=Qwen2.5, Mixing ratio (r)=0.12026.0310.6992.65
AegisModel=Qwen2.5, Mixing ratio (r)=0.12026.038.7292.56
NoneModel=Llama3, Mixing ratio (r)=0.12026.036.8692.33
AegisModel=Llama3, Mixing ratio (r)=0.12026.032.7192.22
AegisModel=OLMo2, Mixing ratio (r)=0.12026.032.2288.57
NoneModel=OLMo2, Mixing ratio (r)=0.12026.030.8188.76
GR-SAPModel=Llama3, Mixing ratio (r)=0.12026.030.7691.62
OriginalModel=OLMo2, Mixing ratio (r)=0.12026.030.4888.45
GR-SAPModel=OLMo2, Mixing ratio (r)=0.12026.030.4788.42
BF16Model Size=1.3B, Precision=BF16, Zero-shot=true2025.01—50.8
BF16Model Size=7B, Precision=BF16, Zero-shot=true2025.01—52.03
BF16Model Size=13B, Precision=BF16, Zero-shot=true2025.01—53.56
FP4Model Size=1.3B, Precision=FP4, Zero-shot=true2025.01—50.82
FP4Model Size=7B, Precision=FP4, Zero-shot=true2025.01—52.97
FP4Model Size=13B, Precision=FP4, Zero-shot=true2025.01—54.12