Research map / Reinforcement learning / Policy gradient and actor-critic
Latent Spherical Flow Policy for Reinforcement Learning with Combinatorial Actions
Lingkai Kong, Anagha Satish, Hezi Jiang, Akseli Kangaslahti, Andrew Ma, Wenbo Chen, Mingxiao Song, Lily Xu, Milind Tambe
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use stochastic latent spherical flow policy instead of deterministic structured policy (SRL) | LSFlow · 4 combinatorial RL tasks | average return: 22.91 -> 27.62 | Improved |
| train flow policy in latent cost space with one solver call per step instead of differentiating through solver (SRL) | LSFlow · 4 combinatorial RL tasks | average training time (h): 3.89 -> 1.29 | Improved |
| add spherical flow matching in place of Euclidean flow (ablation) | LSFlow · Dynamic Scheduling and Dynamic Routing | return | Improved |
| add moderate vMF smoothing of the Bellman operator vs none or too strong (ablation) | LSFlow · combinatorial RL tasks | return | Improved |
| train critic in latent cost space instead of action space (ablation) | LSFlow · Dynamic Scheduling | time per policy update: 6 minutes -> under 1 second | Improved |
| add LSFlow instead of SEQUOIA's MIP-embedded Q-network | LSFlow · Dynamic Scheduling | return: 24.00 -> 28.85 | Improved |
| use LSFlow vs SRL on STI testing networks | LSFlow · STI testing (4 disease networks) | fraction of positives detected | Mixed |
Closest papers by meaning
- Flow Matching for Offline Reinforcement Learning with Discrete Actions (NeurIPS 2026 · similarity 0.88)
- Structured Reinforcement Learning for Combinatorial Decision-Making (NeurIPS 2025 · similarity 0.88)
- Reinforcement learning with combinatorial actions for coupled restless bandits (ICLR 2025 · similarity 0.87)
- Aligning Flow Map Policies with Optimal $Q$-Guidance (NeurIPS 2026 · similarity 0.87)
- Flow-Based Policy for Online Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- Random Policy Evaluation Uncovers Policies of Generative Flow Networks (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
