List your company

Logical Reasoning on LogicBench

80.4AccuracySA-GRPO
47.53656.06864.673.132Sep 5, 2025

Evaluation results

MethodLinks
SA-GRPOBackbone=Qwen3-4B2025.0980.4208
PPOBackbone=Qwen3-4B2025.0980455
GRPOBackbone=Qwen3-4B2025.0978.4416
SA-PPOBackbone=Qwen3-4B2025.0977.6342
GRPO-ERBackbone=Qwen3-4B2025.0977.6263
GRPO-O1Backbone=Qwen3-4B2025.0976350
BaseBackbone=Qwen3-4B2025.0975.4336
GRPOBackbone=Phi-3.5-mini2025.0967.6243
BaseBackbone=Phi-3.5-mini2025.0966309
SA-GRPOBackbone=Qwen3-1.7B2025.0965278
SA-PPOBackbone=Qwen3-1.7B2025.0964.6368
SA-PPOBackbone=Phi-3.5-mini2025.0964.2205
GRPOBackbone=Qwen3-1.7B2025.0964.2372
GRPO-ERBackbone=Qwen3-1.7B2025.0961.4206
PPOBackbone=Phi-3.5-mini2025.0960.8229
GRPO-O1Backbone=Qwen3-1.7B2025.0960.6264
PPOBackbone=Qwen3-1.7B2025.0960.2340
SA-GRPOBackbone=Phi-3.5-mini2025.0957.4173
BaseBackbone=Qwen3-1.7B2025.0956182
SA-PPOBackbone=Gemma3-1B2025.0955.6241
SA-GRPOBackbone=Gemma3-1B2025.0955.2222
GRPOBackbone=Gemma3-1B2025.0953277
GRPO-O1Backbone=Gemma3-1B2025.0952.6223
GRPO-O1Backbone=Phi-3.5-mini2025.0951212
BaseBackbone=Gemma3-1B2025.0950.6303
GRPO-ERBackbone=Gemma3-1B2025.0950250
GRPO-ERBackbone=Phi-3.5-mini2025.0949.8143
PPOBackbone=Gemma3-1B2025.0948.8264