The recent success of denoising diffusion models has significantly advanced text-to-image generation. While these large-scale pretrained models show excellent performance in general image synthesis, downstream objectives often require fine-tuning to meet specific criteria such as aesthetics or human preference. Reward gradient-based strategies are promising in this context, yet existing methods are limited to single-reward tasks, restricting their applicability in real-world scenarios that demand adapting to multiple objectives introduced incrementally over time. In this paper, we first define this more realistic and unexplored problem, termed Reward Incremental Learning (RIL), where models are desired to adapt to multiple downstream objectives incrementally. Additionally, while the models adapt to the ever-emerging new objectives, we observe a unique form of catastrophic forgetting in diffusion model fine-tuning, affecting both metric-wise and visual structure-wise image quality. To address this catastrophic forgetting challenge, we propose Reward Incremental Distillation (RID), a method that mitigates forgetting with minimal computational overhead, enabling stable performance across sequential reward tasks. The experimental results demonstrate the efficacy of RID in achieving consistent, high-quality generation in RIL scenarios. The source code of our work will be publicly available upon acceptance.

Image generation has shown remarkable results in generating high-fidelity realistic images, in particular with the advancement of diffusion-based models. However, the prevalence of AI-generated images may have side effects for the machine learning community that are not clearly identified. Meanwhile, the success of deep learning in computer vision is driven by the massive dataset collected on the Internet. The extensive quantity of synthetic data being added to the Internet would become an obstacle for future researchers to collect "clean" datasets without AI-generated content. Prior research has shown that using datasets contaminated by synthetic images may result in performance degradation when used for training. In this paper, we investigate the potential impact of contaminated datasets on Online Continual Learning (CL) research. We experimentally show that contaminated datasets might hinder the training of existing online CL methods. Also, we propose Entropy Selection with Real-synthetic similarity Maximization (ESRM), a method to alleviate the performance deterioration caused by synthetic images when training online CL models. Experiments show that our method can significantly alleviate performance deterioration, especially when the contamination is severe.

Few-learn learning (FSL) is one of the key future steps in machine learning and has raised a lot of attention. However, in contrast to the rapid development in other domains, such as Computer Vision, the progress of FSL in Nature Language Processing (NLP) is much slower. One of the key reasons for this is the lacking of public benchmarks. NLP FSL researches always report new results on their own constructed few-shot datasets, which is pretty inefficient in results comparison and thus impedes cumulative progress. In this paper, we present FewJoint, a novel Few-Shot Learning benchmark for NLP. Different from most NLP FSL research that only focus on simple N-classification problems, our benchmark introduces few-shot joint dialogue language understanding, which additionally covers the structure prediction and multi-task reliance problems. This allows our benchmark to reflect the real-word NLP complexity beyond simple N-classification. Our benchmark is used in the few-shot learning contest of SMP2020-ECDT task-1. We also provide a compatible FSL platform to ease experiment set-up.