When the training dataset comprises a 1:1 proportion of dogs to cats, a generative model that produces 1:1 dogs and cats better resembles the training species distribution than another model with 3:1 dogs and cats. Can we capture this phenomenon using existing metrics? Unfortunately, we cannot, because these metrics do not provide any interpretability beyond "diversity". In this context, we propose a new evaluation protocol that measures the divergence of a set of generated images from the training set regarding the distribution of attribute strengths as follows. Single-attribute Divergence (SaD) measures the divergence regarding PDFs of a single attribute. Paired-attribute Divergence (PaD) measures the divergence regarding joint PDFs of a pair of attributes. They provide which attributes the models struggle. For measuring the attribute strengths of an image, we propose Heterogeneous CLIPScore (HCS) which measures the cosine similarity between image and text vectors with heterogeneous initial points. With SaD and PaD, we reveal the following about existing generative models. ProjectedGAN generates implausible attribute relationships such as a baby with a beard even though it has competitive scores of existing metrics. Diffusion models struggle to capture diverse colors in the datasets. The larger sampling timesteps of latent diffusion model generate the more minor objects including earrings and necklaces. Stable Diffusion v1.5 better captures the attributes than v2.1. Our metrics lay a foundation for explainable evaluations of generative models.
This paper addresses the challenge of generating Counterfactual Explanations (CEs), involving the identification and modification of the fewest necessary features to alter a classifier's prediction for a given image. Our proposed method, Text-to-Image Models for Counterfactual Explanations (TIME), is a black-box counterfactual technique based on distillation. Unlike previous methods, this approach requires solely the image and its prediction, omitting the need for the classifier's structure, parameters, or gradients. Before generating the counterfactuals, TIME introduces two distinct biases into Stable Diffusion in the form of textual embeddings: the context bias, associated with the image's structure, and the class bias, linked to class-specific features learned by the target classifier. After learning these biases, we find the optimal latent code applying the classifier's predicted class token and regenerate the image using the target embedding as conditioning, producing the counterfactual explanation. Extensive empirical studies validate that TIME can generate explanations of comparable effectiveness even when operating within a black-box setting.
Large-scale text-to-image diffusion models have shown impressive capabilities across various generative tasks, enabled by strong vision-language alignment obtained through pre-training. However, most vision-language discriminative tasks require extensive fine-tuning on carefully-labeled datasets to acquire such alignment, with great cost in time and computing resources. In this work, we explore directly applying a pre-trained generative diffusion model to the challenging discriminative task of visual grounding without any fine-tuning and additional training dataset. Specifically, we propose VGDiffZero, a simple yet effective zero-shot visual grounding framework based on text-to-image diffusion models. We also design a comprehensive region-scoring method considering both global and local contexts of each isolated proposal. Extensive experiments on RefCOCO, RefCOCO+, and RefCOCOg show that VGDiffZero achieves strong performance on zero-shot visual grounding.
We introduce the Qwen-VL series, a set of large-scale vision-language models (LVLMs) designed to perceive and understand both text and images. Comprising Qwen-VL and Qwen-VL-Chat, these models exhibit remarkable performance in tasks like image captioning, question answering, visual localization, and flexible interaction. The evaluation covers a wide range of tasks including zero-shot captioning, visual or document visual question answering, and grounding. We demonstrate the Qwen-VL outperforms existing LVLMs. We present their architecture, training, capabilities, and performance, highlighting their contributions to advancing multimodal artificial intelligence. Code, demo and models are available at https://github.com/QwenLM/Qwen-VL.
Text Style Transfer (TST) is challenging to evaluate because the quality of the generated text manifests itself in multiple aspects, each of which is hard to measure individually: style transfer accuracy, content preservation, and overall fluency of the text. Human evaluation is the gold standard in TST evaluation; however, it is expensive, and the results are difficult to reproduce. Numerous automated metrics are employed to assess performance in these aspects, serving as substitutes for human evaluation. However, the correlation between many of these automated metrics and human evaluations remains unclear, raising doubts about their effectiveness as reliable benchmarks. Recent advancements in Large Language Models (LLMs) have demonstrated their ability to not only match but also surpass the average human performance across a wide range of unseen tasks. This suggests that LLMs have the potential to serve as a viable alternative to human evaluation and other automated metrics. We assess the performance of different LLMs on TST evaluation by employing multiple input prompts and comparing their results. Our findings indicate that (even zero-shot) prompting correlates strongly with human evaluation and often surpasses the performance of (other) automated metrics. Additionally, we propose the ensembling of prompts and show it increases the robustness of TST evaluation.This work contributes to the ongoing efforts in evaluating LLMs on diverse tasks, which includes a discussion of failure cases and limitations.
During the pre-training step of natural language models, the main objective is to learn a general representation of the pre-training dataset, usually requiring large amounts of textual data to capture the complexity and diversity of natural language. Contrasting this, in most cases, the size of the data available to solve the specific downstream task is often dwarfed by the aforementioned pre-training dataset, especially in domains where data is scarce. We introduce controlled randomness, i.e. noise, into the training process to improve fine-tuning language models and explore the performance of targeted noise in addition to the parameters of these models. We find that adding such noise can improve the performance in our two downstream tasks of joint named entity recognition and relation extraction and text summarization.
The ability to detect and track the dynamic objects in different scenes is fundamental to real-world applications, e.g., autonomous driving and robot navigation. However, traditional Multi-Object Tracking (MOT) is limited to tracking objects belonging to the pre-defined closed-set categories. Recently, Open-Vocabulary MOT (OVMOT) and Generic MOT (GMOT) are proposed to track interested objects beyond pre-defined categories with the given text prompt and template image. However, the expensive well pre-trained (vision-)language model and fine-grained category annotations are required to train OVMOT models. In this paper, we focus on GMOT and propose a simple but effective method, Siamese-DETR, for GMOT. Only the commonly used detection datasets (e.g., COCO) are required for training. Different from existing GMOT methods, which train a Single Object Tracking (SOT) based detector to detect interested objects and then apply a data association based MOT tracker to get the trajectories, we leverage the inherent object queries in DETR variants. Specifically: 1) The multi-scale object queries are designed based on the given template image, which are effective for detecting different scales of objects with the same category as the template image; 2) A dynamic matching training strategy is introduced to train Siamese-DETR on commonly used detection datasets, which takes full advantage of provided annotations; 3) The online tracking pipeline is simplified through a tracking-by-query manner by incorporating the tracked boxes in previous frame as additional query boxes. The complex data association is replaced with the much simpler Non-Maximum Suppression (NMS). Extensive experimental results show that Siamese-DETR surpasses existing MOT methods on GMOT-40 dataset by a large margin.
Facial Expression Recognition (FER) is a crucial task in affective computing, but its conventional focus on the seven basic emotions limits its applicability to the complex and expanding emotional spectrum. To address the issue of new and unseen emotions present in dynamic in-the-wild FER, we propose a novel vision-language model that utilises sample-level text descriptions (i.e. captions of the context, expressions or emotional cues) as natural language supervision, aiming to enhance the learning of rich latent representations, for zero-shot classification. To test this, we evaluate using zero-shot classification of the model trained on sample-level descriptions on four popular dynamic FER datasets. Our findings show that this approach yields significant improvements when compared to baseline methods. Specifically, for zero-shot video FER, we outperform CLIP by over 10\% in terms of Weighted Average Recall and 5\% in terms of Unweighted Average Recall on several datasets. Furthermore, we evaluate the representations obtained from the network trained using sample-level descriptions on the downstream task of mental health symptom estimation, achieving performance comparable or superior to state-of-the-art methods and strong agreement with human experts. Namely, we achieve a Pearson's Correlation Coefficient of up to 0.85 on schizophrenia symptom severity estimation, which is comparable to human experts' agreement. The code is publicly available at: https://github.com/NickyFot/EmoCLIP.
We propose task-adaptive tokenization as a way to adapt the generation pipeline to the specifics of a downstream task and enhance long-form generation in mental health. Inspired by insights from cognitive science, our task-adaptive tokenizer samples variable segmentations from multiple outcomes, with sampling probabilities optimized based on task-specific data. We introduce a strategy for building a specialized vocabulary and introduce a vocabulary merging protocol that allows for the integration of task-specific tokens into the pre-trained model's tokenization step. Through extensive experiments on psychological question-answering tasks in both Chinese and English, we find that our task-adaptive tokenization approach brings a significant improvement in generation performance while using up to 60% fewer tokens. Preliminary experiments point to promising results when using our tokenization approach with very large language models.
Free text comments (FTC) in patient-reported outcome measures (PROMs) data are typically analysed using manual methods, such as content analysis, which is labour-intensive and time-consuming. Machine learning analysis methods are largely unsupervised, necessitating post-analysis interpretation. Weakly supervised text classification (WSTC) can be a valuable method of analysis to classify domain-specific text data in which there is limited labelled data. In this paper, we apply five WSTC techniques to FTC in PROMs data to identify health-related quality of life (HRQoL) themes reported by colorectal cancer patients. The WSTC methods label all the themes mentioned in the FTC. The results showed moderate performance on the PROMs data, mainly due to the precision of the models, and variation between themes. Evaluation of the classification performance illustrated the potential and limitations of keyword based WSTC to label PROMs FTC when labelled data is limited.