做多模态RL训练的团队终于有了一个能统一处理文本、图像、视频的框架，不用再为每种模态搭不同的栈。腾讯混元把自家模型验证过的FlowDPPO和DRPO算法也开源了，做扩散模型或LLM RL优化的可以直接拿来用。
腾讯混元开源了UniRL，一个统一的多模态强化学习训练框架。它用一个循环（生成→评分→优势计算→更新→同步）覆盖文本、图像、视频等多种模态，模型和算法作为独立轴，实现模型×算法的组合覆盖。内置FlowDPPO和DRPO两种原创算法，分别针对扩散模型和LLM的强化学习优化。支持可插拔的rollout引擎、FSDP2分片和三种部署模式，旨在解决现有RL栈只能处理单一模态的问题。
1、Most RL stacks are built for one modality. UniRL applies a single post-training loop — generate → ...
1、Most RL stacks are built for one modality. UniRL applies a single post-training loop — generate → score → advantage → update → sync — across model families. Model and algorithm are two independent axes, so your coverage is the model × algorithm product, not a fixed recipe menu. 2、One loop, every modality: text→image, text/image→video, vision-language, text-only LLM and VLM, the LLM→diffusion prompt-enhancer, and unified autoregressive+diffusion generation (Hunyuan-Image 3 and Bagel) — a model class no single-purpose RL repo can even express. 3、Built to scale: pluggable rollout engines (train-side / SGLang / vLLM-Omni) behind one typed contract, FSDP2 sharding, and three deployment modes from a single config knob. 4、Two team-original algorithms headline the release: FlowDPPO: Policy optimization for flow/diffusion models with trust-region masks based on exact divergence (See our paper: Flow-DPPO: Divergence Proximal Policy Optimization for Flow Matching Models github.com/Tencent-Hunyua… ) DRPO: LLM RL with a smooth, advantage-weighted quadratic regularizer (See our paper: Rethinking the Divergence Regularization in LLM RL [ arxiv.org/abs/2606.09821 ]) 💬 0 🔄 0 ❤️ 3 👀 823 📊 1 ⚡