Y. Ma, “Contribution-aware Token Compression for Efficient Video Understanding via Reinforcement Learning”, AAAI, vol. 40, no. 29, pp. 24317–24325, Mar. 2026.