SAM-AI-Reasoning-14B

Official submission of SAM-AI Reasoning Engine trained with Group Relative Policy Optimization (GRPO).

Evaluation

Submitted for independent verification on the Hugging Face Open LLM Leaderboard v2:

• IFEval (Instruction Following)
• BBH (Big Bench Hard)
• MATH-Hard (Advanced Competition Mathematics)
• GPQA Diamond (PhD-level Science)
• MuSR (Multi-step Soft Reasoning)
• MMLU-Pro (Professional Knowledge)