SEE-DPO: Self Entropy Enhanced Direct Preference Optimization

Direct Preference Optimization (DPO) has been successfully used to align large language models (LLMs) according to human preferences, and more recently it has also been applied to improving the quality of text-to-image diffusion models. However, DPO-based methods such as SPO, Diffusion-DPO, and D3PO are highly susceptible to overfitting and reward hacking, especially when the generative model is optimized to fit out-of-distribution during prolonged training. To overcome these challenges and stabilize the training of diffusion models, we introduce a self-entropy regularization mechanism in reinforcement learning from human feedback. This enhancement improves DPO training by encouraging broader exploration and greater robustness. Our regularization technique effectively mitigates reward hacking, leading to improved stability and enhanced image quality across the latent space. Extensive experiments demonstrate that integrating human feedback with self-entropy regularization can significantly boost image diversity and specificity, achieving state-of-the-art results on key image generation metrics.

Self-Ensembling Gaussian Splatting for Few-shot Novel View Synthesis

3D Gaussian Splatting (3DGS) has demonstrated remarkable effectiveness for novel view synthesis (NVS). However, the 3DGS model tends to overfit when trained with sparse posed views, limiting its generalization capacity for broader pose variations. In this paper, we alleviate the overfitting problem by introducing a self-ensembling Gaussian Splatting (SE-GS) approach. We present two Gaussian Splatting models named the $\mathbf{\Sigma}$-model and the $\mathbf{\Delta}$-model. The $\mathbf{\Sigma}$-model serves as the primary model that generates novel-view images during inference. At the training stage, the $\mathbf{\Sigma}$-model is guided away from specific local optima by an uncertainty-aware perturbing strategy. We dynamically perturb the $\mathbf{\Delta}$-model based on the uncertainties of novel-view renderings across different training steps, resulting in diverse temporal models sampled from the Gaussian parameter space without additional training costs. The geometry of the $\mathbf{\Sigma}$-model is regularized by penalizing discrepancies between the $\mathbf{\Sigma}$-model and the temporal samples. Therefore, our SE-GS conducts an effective and efficient regularization across a large number of Gaussian Splatting models, resulting in a robust ensemble, the $\mathbf{\Sigma}$-model. Experimental results on the LLFF, Mip-NeRF360, DTU, and MVImgNet datasets show that our approach improves NVS quality with few-shot training views, outperforming existing state-of-the-art methods. The code is released at https://github.com/sailor-z/SE-GS.

Unlocking Comics: The AI4VA Dataset for Visual Understanding

In the evolving landscape of deep learning, there is a pressing need for more comprehensive datasets capable of training models across multiple modalities. Concurrently, in digital humanities, there is a growing demand to leverage technology for diverse media adaptation and creation, yet limited by sparse datasets due to copyright and stylistic constraints. Addressing this gap, our paper presents a novel dataset comprising Franco-Belgian comics from the 1950s annotated for tasks including depth estimation, semantic segmentation, saliency detection, and character identification. It consists of two distinct and consistent styles and incorporates object concepts and labels taken from natural images. By including such diverse information across styles, this dataset not only holds promise for computational creativity but also offers avenues for the digitization of art and storytelling innovation. This dataset is a crucial component of the AI4VA Workshop Challenges~\url{https://sites.google.com/view/ai4vaeccv2024}, where we specifically explore depth and saliency. Dataset details at \url{https://github.com/IVRL/AI4VA}.

Fully First-Order Methods for Decentralized Bilevel Optimization

This paper focuses on decentralized stochastic bilevel optimization (DSBO) where agents only communicate with their neighbors. We propose Decentralized Stochastic Gradient Descent and Ascent with Gradient Tracking (DSGDA-GT), a novel algorithm that only requires first-order oracles that are much cheaper than second-order oracles widely adopted in existing works. We further provide a finite-time convergence analysis showing that for $n$ agents collaboratively solving the DSBO problem, the sample complexity of finding an $\epsilon$-stationary point in our algorithm is $\mathcal{O}(n^{-1}\epsilon^{-7})$, which matches the currently best-known results of the single-agent counterpart with linear speedup. The numerical experiments demonstrate both the communication and training efficiency of our algorithm.

Bridge-Coder: Unlocking LLMs' Potential to Overcome Language Gaps in Low-Resource Code

Large Language Models (LLMs) demonstrate strong proficiency in generating code for high-resource programming languages (HRPLs) like Python but struggle significantly with low-resource programming languages (LRPLs) such as Racket or D. This performance gap deepens the digital divide, preventing developers using LRPLs from benefiting equally from LLM advancements and reinforcing disparities in innovation within underrepresented programming communities. While generating additional training data for LRPLs is promising, it faces two key challenges: manual annotation is labor-intensive and costly, and LLM-generated LRPL code is often of subpar quality. The underlying cause of this issue is the gap between natural language to programming language gap (NL-PL Gap), which is especially pronounced in LRPLs due to limited aligned data. In this work, we introduce a novel approach called Bridge-Coder, which leverages LLMs' intrinsic capabilities to enhance the performance on LRPLs. Our method consists of two key stages. Bridge Generation, where we create high-quality dataset by utilizing LLMs' general knowledge understanding, proficiency in HRPLs, and in-context learning abilities. Then, we apply the Bridged Alignment, which progressively improves the alignment between NL instructions and LRPLs. Experimental results across multiple LRPLs show that Bridge-Coder significantly enhances model performance, demonstrating the effectiveness and generalization of our approach. Furthermore, we offer a detailed analysis of the key components of our method, providing valuable insights for future work aimed at addressing the challenges associated with LRPLs.

MPDS: A Movie Posters Dataset for Image Generation with Diffusion Model

Movie posters are vital for captivating audiences, conveying themes, and driving market competition in the film industry. While traditional designs are laborious, intelligent generation technology offers efficiency gains and design enhancements. Despite exciting progress in image generation, current models often fall short in producing satisfactory poster results. The primary issue lies in the absence of specialized poster datasets for targeted model training. In this work, we propose a Movie Posters DataSet (MPDS), tailored for text-to-image generation models to revolutionize poster production. As dedicated to posters, MPDS stands out as the first image-text pair dataset to our knowledge, composing of 373k+ image-text pairs and 8k+ actor images (covering 4k+ actors). Detailed poster descriptions, such as movie titles, genres, casts, and synopses, are meticulously organized and standardized based on public movie synopsis, also named movie-synopsis prompt. To bolster poster descriptions as well as reduce differences from movie synopsis, further, we leverage a large-scale vision-language model to automatically produce vision-perceptive prompts for each poster, then perform manual rectification and integration with movie-synopsis prompt. In addition, we introduce a prompt of poster captions to exhibit text elements in posters like actor names and movie titles. For movie poster generation, we develop a multi-condition diffusion framework that takes poster prompt, poster caption, and actor image (for personalization) as inputs, yielding excellent results through the learning of a diffusion model. Experiments demonstrate the valuable role of our proposed MPDS dataset in advancing personalized movie poster generation. MPDS is available at https://anonymous.4open.science/r/MPDS-373k-BD3B.

Personalized Visual Instruction Tuning

Recent advancements in multimodal large language models (MLLMs) have demonstrated significant progress; however, these models exhibit a notable limitation, which we refer to as "face blindness". Specifically, they can engage in general conversations but fail to conduct personalized dialogues targeting at specific individuals. This deficiency hinders the application of MLLMs in personalized settings, such as tailored visual assistants on mobile devices, or domestic robots that need to recognize members of the family. In this paper, we introduce Personalized Visual Instruction Tuning (PVIT), a novel data curation and training framework designed to enable MLLMs to identify target individuals within an image and engage in personalized and coherent dialogues. Our approach involves the development of a sophisticated pipeline that autonomously generates training data containing personalized conversations. This pipeline leverages the capabilities of various visual experts, image generation models, and (multi-modal) large language models. To evaluate the personalized potential of MLLMs, we present a benchmark called P-Bench, which encompasses various question types with different levels of difficulty. The experiments demonstrate a substantial personalized performance enhancement after fine-tuning with our curated dataset.

Rotated Runtime Smooth: Training-Free Activation Smoother for accurate INT4 inference

Large language models have demonstrated promising capabilities upon scaling up parameters. However, serving large language models incurs substantial computation and memory movement costs due to their large scale. Quantization methods have been employed to reduce service costs and latency. Nevertheless, outliers in activations hinder the development of INT4 weight-activation quantization. Existing approaches separate outliers and normal values into two matrices or migrate outliers from activations to weights, suffering from high latency or accuracy degradation. Based on observing activations from large language models, outliers can be classified into channel-wise and spike outliers. In this work, we propose Rotated Runtime Smooth (RRS), a plug-and-play activation smoother for quantization, consisting of Runtime Smooth and the Rotation operation. Runtime Smooth (RS) is introduced to eliminate channel-wise outliers by smoothing activations with channel-wise maximums during runtime. The rotation operation can narrow the gap between spike outliers and normal values, alleviating the effect of victims caused by channel-wise smoothing. The proposed method outperforms the state-of-the-art method in the LLaMA and Qwen families and improves WikiText-2 perplexity from 57.33 to 6.66 for INT4 inference.

From Lists to Emojis: How Format Bias Affects Model Alignment

In this paper, we study format biases in reinforcement learning from human feedback (RLHF). We observe that many widely-used preference models, including human evaluators, GPT-4, and top-ranking models on the RewardBench benchmark, exhibit strong biases towards specific format patterns, such as lists, links, bold text, and emojis. Furthermore, large language models (LLMs) can exploit these biases to achieve higher rankings on popular benchmarks like AlpacaEval and LMSYS Chatbot Arena. One notable example of this is verbosity bias, where current preference models favor longer responses that appear more comprehensive, even when their quality is equal to or lower than shorter, competing responses. However, format biases beyond verbosity remain largely underexplored in the literature. In this work, we extend the study of biases in preference learning beyond the commonly recognized length bias, offering a comprehensive analysis of a wider range of format biases. Additionally, we show that with a small amount of biased data (less than 1%), we can inject significant bias into the reward model. Moreover, these format biases can also be easily exploited by downstream alignment algorithms, such as best-of-n sampling and online iterative DPO, as it is usually easier to manipulate the format than to improve the quality of responses. Our findings emphasize the need to disentangle format and content both for designing alignment algorithms and evaluating models.

Data Augmentation via Latent Diffusion for Saliency Prediction

Saliency prediction models are constrained by the limited diversity and quantity of labeled data. Standard data augmentation techniques such as rotating and cropping alter scene composition, affecting saliency. We propose a novel data augmentation method for deep saliency prediction that edits natural images while preserving the complexity and variability of real-world scenes. Since saliency depends on high-level and low-level features, our approach involves learning both by incorporating photometric and semantic attributes such as color, contrast, brightness, and class. To that end, we introduce a saliency-guided cross-attention mechanism that enables targeted edits on the photometric properties, thereby enhancing saliency within specific image regions. Experimental results show that our data augmentation method consistently improves the performance of various saliency models. Moreover, leveraging the augmentation features for saliency prediction yields superior performance on publicly available saliency benchmarks. Our predictions align closely with human visual attention patterns in the edited images, as validated by a user study.