Tencent Proposes Adaptive Reward Routing for Joint Audio-Video Diffusion RL

Tencent's new paper introduces Adaptive Reward Routing, which dynamically adapts update placement and reward composition during forward-process RL to resolve competing reward objectives when training joint audio-video diffusion models.

2026-10-02 ~ 2026-10-03 · 2 related posts

• Tencent's Adaptive Reward Routing Dynamically Balances Multi-Reward RL for Joint Audio-Video Diffusion — tencent · 2026-10-02
• Adaptive Reward Routing improves joint audio-video diffusion via forward-process RL — _akhaliq · 2026-10-03