Getting Space Right: Agentic Rewards for Image Generation
Reinforcement learning (RL) is promising for spatial text-to-image generation, but existing work largely focuses on policy optimization, assuming accurate rewards and effective online rollouts.
ProofPaper ↗
Key points
- We observe in-domain gains with out-of-domain degradation, exposing two limitations: unreliable spatial rewards in open-world settings and rollouts lacking basic-to-complex progression.
- We propose a curriculum-guided RL framework.
- First, we introduce SpatialCritic, an agentic reward model equipped with a spatial perception harness that combines reusable verification guidance with specialized vision tools.
- We then construct spatial preference datasets through targeted counterfactual perturbations and perform agent-harness alignment to improve tool selection and evidence integration efficiently.
Sources (1)
- [1]Getting Space Right: Agentic Rewards for Image GenerationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 10, 12:13 PM
Reinforcement learning (RL) is promising for spatial text-to-image generation, but existing work largely focuses on policy optimization, assuming accurate rewards and effective online rollouts.
We observe in-domain gains with out-of-domain degradation, exposing two limitations: unreliable spatial rewards in open-world settings and rollouts lacking basic-to-complex progression.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 10, 2026SAGE: Pushing 2-Bit Vector-Quantized LLMs toward Real-World Usability
- Oct 10, 2026Red-Teaming of Vision-Language-Action Models via Quality Diversity Environment Generation
- Oct 10, 2026LightVAE: Towards Compact and Efficient Video Autoencoders
- Oct 9, 2026SAT-RL: Training Language Models for Logical Reasoning with Symbolic Solvers
- Oct 9, 2026Latent World Action Model for GUI Agents
- Oct 9, 2026Rethinking Relative Policy Differences Transfer of On-Policy Distillation