Small RL Controller, Large Language Model: RL-Guided Adaptive Sampling for Test-Time Scaling
Open paper
Benchmark
Model
Rank
Results
mathematical-reasoning-on-aime-2025
RL-guided (Qwen3-4B-Instruct-Thinking)
–
Accuracy: 57.1
mathematical-reasoning-on-aime24
RL-guided (Qwen3-4B-Instruct-Thinking)
–
Acc: 73.0
Rank counts only results with a code link.