Research map / Image generation / Text-to-image diffusion
Talk2Move: Reinforcement Learning for Text-Instructed Object-Level Geometric Transformation in Scenes
Jing Tan, Zhaoyang Zhang, Yantao Shen, Jiarui Cai, Shuo Yang, Jiajun Wu, Wei Xia, Zhuowen Tu, Stefano Soatto
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 |
|---|---|---|---|
| add RL (GRPO) on top of SFT (ablation) | Flux-Kontext+SFT · translation | accuracy: 38.81% -> 47.14% | Improved |
| add RL (GRPO) on top of SFT (ablation) | QwenImageEdit+SFT · translation | accuracy: 67.14% -> 73.13% | Improved |
| add RL on top of SFT trained with 1/10 data (ablation) | QwenImageEdit+SFT · translation | accuracy: 26.67% -> 73.33% | Improved |
| add SFT to the base editor (ablation) | QwenImageEdit · translation | accuracy: 32.86% -> 67.14% | Improved |
| use active step sampling instead of full sampling (ablation) | Talk2Move · translation | accuracy: 69.12% -> 76.67% | Improved |
| use active step sampling instead of sliding window (ablation) | Talk2Move · translation | accuracy: 67.14% -> 76.67% | Improved |
| use active step sampling instead of full sampling (ablation) | Talk2Move · translation | iteration time: 172.32 -> 87.27 | Improved |
| use active step sampling instead of sliding window (ablation) | Talk2Move · translation | iteration time: 101.61 -> 87.27 | Improved |
| use spatial reward instead of VLM reward (ablation) | Talk2Move · rotation | accuracy: 11.63% -> 29.55% | Improved |
| use spatial reward instead of VLM reward (ablation) | Talk2Move · rotation | rotation error: 0.3294 -> 0.2861 | Improved |
| train Talk2Move vs QwenImageEdit base | QwenImageEdit · resize synthetic | scale error: 0.4203 -> 0.3894 | Mixed |
Closest papers by meaning
- MetaSpatial: Reinforcing 3D Spatial Reasoning in VLMs for the Metaverse (ICLR 2026 · similarity 0.88)
- STAR-R1: Multi-View Spatial TrAnsformation Reasoning by Reinforcing Multimodal LLMs (CVPR 2026 · similarity 0.87)
- Multi-Region Text-Driven Manipulation of Diffusion Imagery (AAAI 2024 · similarity 0.87)
- Socratic-Geo: Synthetic Data Generation and Cross-Modal Geometric Reasoning via Multi-Agent Interaction (CVPR 2026 · similarity 0.87)
- Exploring Spatial Intelligence from a Generative Perspective (CVPR 2026 · similarity 0.86)
- Let Language Constrain Geometry: Vision–Language Models as Semantic and Spatial Critics for 3D Generation (ICML 2026 · 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.
