Research map / Reinforcement learning / Visual and representation RL
CellFluxRL: Biologically-Constrained Virtual Cell Modeling via Reinforcement Learning
Dongxia Wu, Shiye Su, Yuhui Zhang, Elaine Sui, Emma Lundberg, Emily Fox, Serena Yeung-Levy
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 |
|---|---|---|---|
| RL post-train CellFlux with 7 biological rewards (CellFluxRL) | CellFlux · BBBC021 | MoA reward: 0.26 -> 0.34 | Improved |
| RL post-train CellFlux with 7 biological rewards (CellFluxRL) | CellFlux · BBBC021 | Overall reward: -2.44 -> 0.46 (Table 1); text says -4.44 -> -1.54 | Mixed |
| RL post-train CellFlux with 7 biological rewards (CellFluxRL) | CellFlux · BBBC021 | Nuc-in-Cyto reward: 0.88 -> 0.96 | Improved |
| RL post-train CellFlux with 7 biological rewards (CellFluxRL) | CellFlux · BBBC021 | FID: 20.36 -> 24.01 | Hurt |
| RL post-train CellFlux with 7 biological rewards (CellFluxRL) | CellFlux · BBBC021 | KID: 0.015 -> 0.014 | Improved |
| add best-of-N test-time scaling (N=4) with the reward | CellFluxRL · BBBC021 | MoA reward: 0.34 -> 0.56 | Improved |
| add best-of-N test-time scaling (N=4) with the reward | CellFluxRL · BBBC021 | FID: 24.01 -> 23.19 | Improved |
| add best-of-N test-time scaling (N=4) with the reward | CellFluxRL · BBBC021 | KID: 0.014 -> 0.016 | Hurt |
| increase number of best-of-N candidates (ablation) | CellFlux and CellFluxRL · BBBC021 | individual rewards | Improved |
| optimize a single reward only instead of the weighted sum (ablation) | CellFluxRL · BBBC021 | rewards other than the targeted one: e.g. MoA-only on NucSize: -3.319 vs -1.038 combined | Mixed |
| combine all seven rewards instead of single-reward training (ablation) | CellFluxRL · BBBC021 | MoA reward: 0.412 (MoA only) vs 0.337 (combined) | Mixed |
| vary KL weight beta from 1.0 to 1.3 (ablation) | CellFluxRL · BBBC021 | structural/morphological vs biological rewards | Mixed |
Closest papers by meaning
- CellFlux: Simulating Cellular Morphology Changes via Flow Matching (ICML 2025 · similarity 0.88)
- Feedback Efficient Online Fine-Tuning of Diffusion Models (ICML 2024 · similarity 0.85)
- Reinforcement Learning for Control of Non-Markovian Cellular Population Dynamics (ICLR 2025 · similarity 0.85)
- Bridging Model-Based Optimization and Generative Modeling via Conservative Fine-Tuning of Diffusion Models (NeurIPS 2024 · similarity 0.84)
- CellCLIP - Learning Perturbation Effects in Cell Painting via Text-Guided Contrastive Learning (NeurIPS 2025 · similarity 0.84)
- TFG-Flow: Training-free Guidance in Multimodal Generative Flow (ICLR 2025 · similarity 0.83)
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.
