Our new X account is live! Follow @wizwand_team for updates

Social Commonsense Reasoning on SIQA

85.24Accuracy

In-Squeeze

Updated 4d ago

Evaluation Results

Method	Links
In-Squeeze 2026.02		85.24
In-Squeeze 2026.02		84.41
Direct Fine-tuning 2026.02		84.25
Cont-Squeeze 2026.02		84.25
Direct Fine-tuning 2026.02		83.68
Cont-Squeeze 2026.02		83.68
Direct Fine-tuning 2026.02		83.16
Qwen3-4B 2025.12		75.59
S-MeZO 2024.02		70.2
Qwen2.5-3B 2025.12		69.4
Qwen3-1.7B 2025.12		68.58
MeZO 2024.02		68.5
gemma2-2B 2025.12		65.92
SmolLM3-3B 2025.12		65.25
Qwen2.5-1.5B 2025.12		64.94
llama-3.2-3B 2025.12		64.33
Qwen2-1.5B 2025.12		63.46
YuLan-Mini-2.4B 2025.12		63.25
PCMind-2.1-Kaiyuan-2B 2025.12		62.59
Qwen3-0.6B 2025.12		61.51
SmolLM2-1.7B 2025.12		60.18
OLMo-2-0425-1B 2025.12		55.53
llama-3.2-1B 2025.12		50.61
ShapeLLM-Omni-7B 2025.12		41.5
CoRe3D 2025.12		41.5
Qwen2.5-vl-7B 2025.12		41
LLaMA3.2-Vision-11B 2025.12		40.6
LLaMA-Mesh-8B 2025.12		40.3
DLM + extra token 2026.01		37.92
DLM 2026.01		35.98
DLM + GA 2026.01		34.44
Cont-Squeeze 2026.02		34