Research map / Image generation / Text-to-image diffusion
RewardFlow: Generate Images by Optimizing What You Reward
Onkar Susladkar, Dong-Hwan Jang, Tushar Prakash, Adheesh Juvekar, Vedant Shah, Ayush Barik, Nabeel Bashir, Muntasir Wahed, Ritish Shrirao, Ismini Lourentzou
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 Dynamic Reward Weighting (ablation) | RewardFlow · PIE-Bench (Qwen Image) | PSNR: 30.77 -> 32.09 | Improved |
| add Semantic Primitives (ablation) | RewardFlow · PIE-Bench (Qwen Image) | Distance: 9.03 -> 7.64 | Improved |
| add Reward-Aware Step Size (ablation) | RewardFlow · PIE-Bench (Qwen Image) | Distance: 9.15 -> 7.64 | Improved |
| add KL tether (ablation) | RewardFlow · PIE-Bench (Qwen Image) | PSNR: 29.98 -> 32.09 | Improved |
| add perceptual reward to global reward (ablation) | RewardFlow · PIE-Bench (Qwen Image) | LPIPS: 45.22 -> 43.12 | Improved |
| add human preference reward (ablation) | RewardFlow · PIE-Bench (Qwen Image) | PSNR: 27.57 -> 28.82 | Improved |
| add object consistency reward (ablation) | RewardFlow · PIE-Bench (Qwen Image) | Distance: 9.77 -> 8.39 | Improved |
| add region-level grounding reward (ablation) | RewardFlow · PIE-Bench (Qwen Image) | Whole CLIP score: 21.47 -> 26.47 | Improved |
| add VQA reward (ablation) | RewardFlow · PIE-Bench (Qwen Image) | PSNR: 30.12 -> 32.09 | Improved |
| add RewardFlow guidance to Flux | Flux · T2I-CompBench Complex | accuracy: 0.47 -> 0.57 | Improved |
| add RewardFlow guidance to Qwen Image | Qwen Image · T2I-CompBench Complex | accuracy: 0.61 -> 0.78 | Improved |
| add RewardFlow guidance to PixArt-alpha | PixArt-alpha · T2I-CompBench Complex | accuracy: 0.41 -> 0.52 | Improved |
Later papers that compared with this method
1 later accepted paper reports RewardFlow as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- Edit the Bits, Diff the Codes: Bitwise Residual Editing for Visual Autoregressive Models (NeurIPS 2026 · better on 0 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- InstantEdit — InstantEdit: Text-Guided Few-Step Image Editing with Piecewise Rectified Flow (ICCV 2025) · better on 6 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Reward-Instruct: A Reward-Centric Approach to Fast Photo-Realistic Image Generation (NeurIPS 2025 · similarity 0.87)
- Learning to Sample Effective and Diverse Prompts for Text-to-Image Generation (CVPR 2025 · similarity 0.87)
- SpatialFlow-GRPO: Where Spatial Credit Drives Image Editing (NeurIPS 2026 · similarity 0.87)
- SplitFlow: Flow Decomposition for Inversion-Free Text-to-Image Editing (NeurIPS 2025 · similarity 0.87)
- Diff-Instruct with Diffused Reward: Towards Principled One-step Generator RL (NeurIPS 2026 · similarity 0.87)
- Multi-Reward as Condition for Instruction-based Image Editing (ICLR 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.
