The Alignment Illusion: How BadReward Corrupts Text-to-Image Generation
cs.LG, cs.AI, cs.CR
Submitted: 2025-06-03
Updated: 2026-11-02
Code: https://github.com/akashsonowal/ddpo-pytorch
Terminology
Sources
- Best-of-Venom: Attacking RLHF by Injecting Poisoned Preference Data
- Poisoning Attacks against Support Vector Machines
- Training Diffusion Models with Reinforcement Learning
- Aligning Text-to-Image Models using Human Feedback
- Backdooring Bias ($B^2$) into Stable Diffusion Models
- Universal Jailbreak Backdoors from Poisoned Human Feedback
- Preference Poisoning Attacks on Reward Model Learning
- Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis
- Shadowcast: Stealthy Data Poisoning Attacks Against Vision-Language Models
- Text-to-image Diffusion Models in Generative AI: A Survey
- Aligning Few-Step Diffusion Models with Dense Reward Difference Learning
- Diffusion Models for Reinforcement Learning: A Survey
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis