List your company

Commonsense Reasoning on Commonsense 8 Sub-Tasks

88.22Accuracy (8 Sub-Tasks)LR-LoRA
38.071251.090664.1177.1294Mar 13, 2026Mar 26, 2026Apr 9, 2026Apr 23, 2026May 6, 2026May 20, 2026Jun 3, 2026

Evaluation results

MethodLinks
LR-LoRABackbone=LLaMA3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0688.22—
LR-LoRABackbone=Phi-3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0687.3—
ElaLoRABackbone=LLaMA3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0687.05—
ElaLoRABackbone=Phi-3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0686.2—
LR-LoRABackbone=LLaMA3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0685.99—
AdaLoRABackbone=LLaMA3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0685.88—
RandLoRABackbone=LLaMA3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0685.59—
SineLoRABackbone=LLaMA3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0685.42—
DyLoRABackbone=LLaMA3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0685.4—
AdaLoRABackbone=Phi-3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0685.35—
LoRABackbone=LLaMA3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0685.24—
RandLoRABackbone=Phi-3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0685.22—
DyLoRABackbone=Phi-3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0685.05—
SineLoRABackbone=Phi-3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0685.01—
LoRABackbone=Phi-3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0684.96—
ElaLoRABackbone=LLaMA3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0684.55—
LR-LoRABackbone=Phi-3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0683.9—
AdaLoRABackbone=LLaMA3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0683.5—
DyLoRABackbone=LLaMA3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0683.25—
LoRABackbone=LLaMA3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0683.09—
ElaLoRABackbone=Phi-3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0682.55—
NoLABackbone=Phi-3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0682.33—
RandLoRABackbone=Phi-3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0682.33—
SineLoRABackbone=LLaMA3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0681.9—
VeRABackbone=LLaMA3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0681.65—
AdaLoRABackbone=Phi-3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0681.45—
VeRABackbone=Phi-3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0681.4—
RandLoRABackbone=LLaMA3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0681.31—
NoLABackbone=LLaMA3-8B, Data Regime (Samples)=170k, Rank (r)=322026.0681.16—
SineLoRABackbone=Phi-3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0681.12—
DyLoRABackbone=Phi-3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0680.85—
NoLABackbone=Phi-3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0680.35—
LoRABackbone=Phi-3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0680.34—
VeRABackbone=Phi-3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0678.63—
VeRABackbone=LLaMA3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0677.14—
NoLABackbone=LLaMA3-8B, Data Regime (Samples)=15k, Rank (r)=322026.0676.87—
LoRARatio=0%, # Params=0.45B2026.0361.455.5
LR-LoRABackbone=Qwen2-0.5B, Data Regime (Samples)=170k, Rank (r)=322026.0661.05—
OriginalRatio=0%, # Params=-2026.0360.637.7
ElaLoRABackbone=Qwen2-0.5B, Data Regime (Samples)=170k, Rank (r)=322026.0659.15—
MoBERatio=30%, # Params=0.45B2026.0358.350.4
AdaLoRABackbone=Qwen2-0.5B, Data Regime (Samples)=170k, Rank (r)=322026.0658.2—
RandLoRABackbone=Qwen2-0.5B, Data Regime (Samples)=170k, Rank (r)=322026.0657.86—
DyLoRABackbone=Qwen2-0.5B, Data Regime (Samples)=170k, Rank (r)=322026.0657.6—
SineLoRABackbone=Qwen2-0.5B, Data Regime (Samples)=170k, Rank (r)=322026.0657.55—
LoRABackbone=Qwen2-0.5B, Data Regime (Samples)=170k, Rank (r)=322026.0657.34—
Replace (w. shared)Ratio=30%, # Params=0.45B2026.0356.752.3
LightMoERatio=30%, # Params=0.45B2026.0356.355.3
LR-LoRABackbone=Qwen2-0.5B, Data Regime (Samples)=15k, Rank (r)=322026.0655.88—
Replace (w/o shared)Ratio=30%, # Params=0.45B2026.0355.652.6
HC-SMoERatio=30%, # Params=0.45B2026.0355.546.8
MC-SMoERatio=30%, # Params=0.45B2026.0355.153.4
Interdomain AttentionZero-shot=true, Model Size=1.3B, Evaluation Harness=lm-evaluation-harness2026.0554.54—
Full FTRatio=0%, # Params=6.92B2026.0354.256.3
ElaLoRABackbone=Qwen2-0.5B, Data Regime (Samples)=15k, Rank (r)=322026.0654.02—
MoBERatio=40%, # Params=0.45B2026.035443
Replace (w. shared)Ratio=40%, # Params=0.45B2026.0353.949.3
LightMoERatio=40%, # Params=0.45B2026.0353.253
AdaLoRABackbone=Qwen2-0.5B, Data Regime (Samples)=15k, Rank (r)=322026.0653.15—
RandLoRABackbone=Qwen2-0.5B, Data Regime (Samples)=15k, Rank (r)=322026.0652.89—
DyLoRABackbone=Qwen2-0.5B, Data Regime (Samples)=15k, Rank (r)=322026.0652.75—
SineLoRABackbone=Qwen2-0.5B, Data Regime (Samples)=15k, Rank (r)=322026.0652.51—
Replace (w/o shared)Ratio=40%, # Params=0.45B2026.0352.550.1
LoRABackbone=Qwen2-0.5B, Data Regime (Samples)=15k, Rank (r)=322026.0652.34—
VeRABackbone=Qwen2-0.5B, Data Regime (Samples)=170k, Rank (r)=322026.0651.78—
SoftmaxZero-shot=true, Model Size=1.3B, Evaluation Harness=lm-evaluation-harness2026.0551.51—
HC-SMoERatio=40%, # Params=0.45B2026.0351.538.4
MC-SMoERatio=40%, # Params=0.45B2026.0350.447.3
S4D-onlyZero-shot=true, Model Size=1.3B, Evaluation Harness=lm-evaluation-harness2026.0549.44—
Replace (w. shared)Ratio=50%, # Params=0.45B2026.0349.341.6
VeRABackbone=Qwen2-0.5B, Data Regime (Samples)=15k, Rank (r)=322026.0648.12—
LightMoERatio=50%, # Params=0.45B2026.0348.148.1
HC-SMoERatio=50%, # Params=0.45B2026.0347.834
NoLABackbone=Qwen2-0.5B, Data Regime (Samples)=170k, Rank (r)=322026.0647.4—
Replace (w/o shared)Ratio=50%, # Params=0.45B2026.0347.144.3
MoBERatio=50%, # Params=0.45B2026.0346.429.9
MC-SMoERatio=50%, # Params=0.45B2026.0344.842.5
NoLABackbone=Qwen2-0.5B, Data Regime (Samples)=15k, Rank (r)=322026.0642.55—
MC-SMoE*Ratio=40%, # Params=1.65B2026.0340.350.2
MC-SMoE*Ratio=50%, # Params=1.65B2026.034045.3