TILT: Model-Intrinsic Reward Alignment For Compositional Diffusion

arXiv:2607.21606 · cs.AI · Submitted 2026-05-16 · Read on arXiv

cs.AI

Submitted: 2026-05-16

Updated: 2026-10-05

Comments: Extended version. An earlier short version appeared in ICML2026 SPIGM workshop

Project page: https://debottam-dutta7.github.io/tilt_web

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Alignment — 3× in this paper · explained in 1 episode
• Hybrid Approach — 1× in this paper · explained in 4 episodes
• Training-Free Framework — 1× in this paper · explained in 2 episodes

Sources

• Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets
• Attend-and-Excite: Attention-Based Semantic Guidance for Text-to-Image Diffusion Models
• Reduce, Reuse, Recycle: Compositional Generation with Energy-Based Diffusion Models and MCMC
• GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment
• Denoising Diffusion Probabilistic Models
• Imagen Video: High Definition Video Generation with Diffusion Models
• Read It Back: Pretrained MLLMs Are Zero-Shot Reward Models for Text-to-Image Generation
• SCORE: Scaling audio generation using Standardized COmposite REwards
• TweedieMix: Improving Multi-Concept Fusion for Diffusion-based Image/Video Generation
• Evaluating Text-to-Visual Generation with Image-to-Text Generation
• T2I-Adapter: Learning Adapters to Dig out More Controllable Ability for Text-to-Image Diffusion Models
• SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis
• Inference-Time Alignment of Diffusion Models with Direct Noise Optimization
• InstanceDiffusion: Instance-level Control for Image Generation
• BoxDiff: Text-to-Image Synthesis with Training-Free Box-Constrained Diffusion
• LoCo: Locally Constrained Training-Free Layout-to-Image Synthesis

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems