Text-to-Image (T2I) models have made remarkable progress in generating images from text prompts, but their output quality and safety still depend heavily on how prompts are phrased. Existing safety methods typically refine prompts using large language models (LLMs), but they overlook the images produced, which can result in unsafe outputs or unnecessary changes to already safe prompts. To address this, we propose an iterative prompt refinement algorithm that uses Vision Language Models (VLMs) to analyze both the input prompts and the generated images. By leveraging visual feedback, our method refines prompts more effectively, improving safety while maintaining user intent and reliability comparable to existing LLM-based approaches. Additionally, we introduce a new dataset labeled with both textual and visual safety signals using off-the-shelf multi-modal LLM, enabling supervised fine-tuning. Experimental results demonstrate that our approach produces safer outputs without compromising alignment with user intent, offering a practical solution for generating safer T2I content. textbftextcolorredWARNING: This paper contains examples of harmful or inappropriate images generated by models.

Iterative Prompt Refinement for Safer Text-to-Image Generation

Direct Preference Optimization (DPO) is a simple and efficient framework that has attracted substantial attention. However, it often struggles to meet its primary objectives---increasing the generation probability of chosen responses while reducing that of rejected responses---due to the dominant influence of rejected responses on the loss function. This imbalance leads to suboptimal performance in promoting preferred responses. In this work, we systematically analyze the limitations of DPO and existing algorithms designed to achieve the objectives stated above. To address these limitations, we propose Bounded-DPO (BDPO), a novel method that bounds the influence of rejected responses while maintaining the original optimization structure of DPO. Through theoretical analysis and empirical evaluations, we demonstrate that BDPO achieves a balanced optimization of the chosen and rejected responses, outperforming existing algorithms.

Rethinking DPO: The Role of Rejected Responses in Preference Misalignment

Language detoxification involves removing toxicity from offensive language. While a neutral-toxic paired dataset provides a straightforward approach for training detoxification models, creating such datasets presents several challenges: i) the need for human annotation to build paired data, and ii) the rapid evolution of offensive terms, rendering static datasets quickly outdated. To tackle these challenges, we introduce an automated paired data generation pipeline, called K/DA. This pipeline is designed to generate offensive language with implicit offensiveness and trend-aligned slang, making the resulting dataset suitable for detoxification model training. We demonstrate that the dataset generated by K/DA exhibits high pair consistency and greater implicit offensiveness compared to existing Korean datasets, and also demonstrates applicability to other languages. Furthermore, it enables effective training of a high-performing detoxification model with simple instruction fine-tuning.

K/DA: Automated Data Generation Pipeline for Detoxifying Implicitly Offensive Language in Korean

One of the major challenges of offline reinforcement learning (RL) is dealing with distribution shifts that stem from the mismatch between the trained policy and the data collection policy. Stationary distribution correction estimation algorithms (DICE) have addressed this issue by regularizing the policy optimization with f-divergence between the state-action visitation distributions of the data collection policy and the optimized policy. While such regularization naturally integrates to derive an objective to get optimal state-action visitation, such an implicit policy optimization framework has shown limited performance in practice. We observe that the reduced performance is attributed to the biased estimate and the properties of conjugate functions of f-divergence regularization. In this paper, we improve the regularized implicit policy optimization framework by relieving the bias and reshaping the conjugate function by relaxing the constraints. We show that the relaxation adjusts the degree of involvement of the sub-optimal samples in optimization, and we derive a new offline RL algorithm that benefits from the relaxed framework, improving from a previous implicit policy optimization algorithm by a large margin.

Relaxed Stationary Distribution Correction Estimation for Improved Offline Policy Optimization

Reinforcement learning (RL) offers a principled framework for optimizing a given reward function and has been applied to a neural machine translation (NMT) problem to maximize arbitrary task metrics. However, previously adopted RL algorithms for NMT (e.g., policy gradient) are generally slow as they require online data collection, and limits the algorithm's applicability to specific reward functions that can be evaluated online.
In this paper, we present an offline RL algorithm called CER (conservative expectile regression). Despite the demanding nature of offline RL tasks, which are even more difficult with large models, this algorithm is capable to learn stably by explicitly exploiting the properties of NMT's RL formulation, such as the deterministic transition function. We analyze and discuss the design choices of CER, and demonstrate in the experiments that the proposed method outperforms its competitors for offline reward optimization in NMT.

Byung-jun Lee

5

19

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES