Spectral-Driven Adaptive-Rank Optimizer for Large Language Model Reinforcement Learning with Verifiable Rewards

Open paper

Benchmark	Model	Rank	Results
mathematical-reasoning-on-minerva-math	Ares (Qwen3-8B-Base)	#38	Accuracy: 29.4
mathematical-reasoning-on-olympiadbench	Ares (Qwen3-8B-Base)	#40	Accuracy: 43.4