H. Wang, “TOP-RL: Task-Optimized Progressive Token Pruning with Reinforcement Learning for Vision Language Models”, AAAI, vol. 40, no. 18, pp. 15824–15832, Mar. 2026.