As Large Language Models (LLMs) are deployed more widely, customization with respect to vocabulary, style and character becomes more important. In this work we introduce model arithmetic, a novel inference framework for composing and biasing LLMs without the need for model (re)training or highly specific datasets. In addition, the framework allows for more precise control of generated text than direct prompting and prior controlled text generation (CTG) techniques. Using model arithmetic, we can express prior CTG techniques as simple formulas and naturally extend them to new and more effective formulations. Further, we show that speculative sampling, a technique for efficient LLM sampling, extends to our setting. This enables highly efficient text generation with multiple composed models with only marginal overhead over a single model. Our empirical evaluation demonstrates that model arithmetic allows fine-grained control of generated text while outperforming state-of-the-art on the task of toxicity reduction.
Understanding human actions from videos of first-person view poses significant challenges. Most prior approaches explore representation learning on egocentric videos only, while overlooking the potential benefit of exploiting existing large-scale third-person videos. In this paper, (1) we develop EgoInstructor, a retrieval-augmented multimodal captioning model that automatically retrieves semantically relevant third-person instructional videos to enhance the video captioning of egocentric videos. (2) For training the cross-view retrieval module, we devise an automatic pipeline to discover ego-exo video pairs from distinct large-scale egocentric and exocentric datasets. (3) We train the cross-view retrieval module with a novel EgoExoNCE loss that pulls egocentric and exocentric video features closer by aligning them to shared text features that describe similar actions. (4) Through extensive experiments, our cross-view retrieval module demonstrates superior performance across seven benchmarks. Regarding egocentric video captioning, EgoInstructor exhibits significant improvements by leveraging third-person videos as references.
While AI-generated text and 2D images continue to expand its territory, 3D generation has gradually emerged as a trend that cannot be ignored. Since the year 2023 an abundant amount of research papers has emerged in the domain of 3D generation. This growth encompasses not just the creation of 3D objects, but also the rapid development of 3D character and motion generation. Several key factors contribute to this progress. The enhanced fidelity in stable diffusion, coupled with control methods that ensure multi-view consistency, and realistic human models like SMPL-X, contribute synergistically to the production of 3D models with remarkable consistency and near-realistic appearances. The advancements in neural network-based 3D storing and rendering models, such as Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS), have accelerated the efficiency and realism of neural rendered models. Furthermore, the multimodality capabilities of large language models have enabled language inputs to transcend into human motion outputs. This paper aims to provide a comprehensive overview and summary of the relevant papers published mostly during the latter half year of 2023. It will begin by discussing the AI generated object models in 3D, followed by the generated 3D human models, and finally, the generated 3D human motions, culminating in a conclusive summary and a vision for the future.
The recent advancements in text-to-3D generation mark a significant milestone in generative models, unlocking new possibilities for creating imaginative 3D assets across various real-world scenarios. While recent advancements in text-to-3D generation have shown promise, they often fall short in rendering detailed and high-quality 3D models. This problem is especially prevalent as many methods base themselves on Score Distillation Sampling (SDS). This paper identifies a notable deficiency in SDS, that it brings inconsistent and low-quality updating direction for the 3D model, causing the over-smoothing effect. To address this, we propose a novel approach called Interval Score Matching (ISM). ISM employs deterministic diffusing trajectories and utilizes interval-based score matching to counteract over-smoothing. Furthermore, we incorporate 3D Gaussian Splatting into our text-to-3D generation pipeline. Extensive experiments show that our model largely outperforms the state-of-the-art in quality and training efficiency.
Recently, text-guided digital portrait editing has attracted more and more attentions. However, existing methods still struggle to maintain consistency across time, expression, and view or require specific data prerequisites. To solve these challenging problems, we propose CosAvatar, a high-quality and user-friendly framework for portrait tuning. With only monocular video and text instructions as input, we can produce animatable portraits with both temporal and 3D consistency. Different from methods that directly edit in the 2D domain, we employ a dynamic NeRF-based 3D portrait representation to model both the head and torso. We alternate between editing the video frames' dataset and updating the underlying 3D portrait until the edited frames reach 3D consistency. Additionally, we integrate the semantic portrait priors to enhance the edited results, allowing precise modifications in specified semantic areas. Extensive results demonstrate that our proposed method can not only accurately edit portrait styles or local attributes based on text instructions but also support expressive animation driven by a source video.
There are many recent advanced developments for the Text-to-SQL task, where the Picard model is one of the the top performing models as measured by the Spider dataset competition. However, bringing Text-to-SQL systems to realistic use-cases through domain adaptation remains a tough challenge. We analyze how well the base T5 Language Model and Picard perform on query structures different from the Spider dataset, we fine-tuned the base model on the Spider data and on independent databases (DB). To avoid accessing the DB content online during inference, we also present an alternative way to disambiguate the values in an input question using a rule-based approach that relies on an intermediate representation of the semantic concepts of an input question. In our results we show in what cases T5 and Picard can deliver good performance, we share the lessons learned, and discuss current domain adaptation challenges.
Scene text image super-resolution (STISR) aims at simultaneously increasing the resolution and readability of low-resolution scene text images, thus boosting the performance of the downstream recognition task. Two factors in scene text images, semantic information and visual structure, affect the recognition performance significantly. To mitigate the effects from these factors, this paper proposes a Prior-Enhanced Attention Network (PEAN). Specifically, a diffusion-based module is developed to enhance the text prior, hence offering better guidance for the SR network to generate SR images with higher semantic accuracy. Meanwhile, the proposed PEAN leverages an attention-based modulation module to understand scene text images by neatly perceiving the local and global dependence of images, despite the shape of the text. A multi-task learning paradigm is employed to optimize the network, enabling the model to generate legible SR images. As a result, PEAN establishes new SOTA results on the TextZoom benchmark. Experiments are also conducted to analyze the importance of the enhanced text prior as a means of improving the performance of the SR network. Code will be made available at https://github.com/jdfxzzy/PEAN.
The ability to provide fine-grained control for generating and editing visual imagery has profound implications for computer vision and its applications. Previous works have explored extending controllability in two directions: instruction tuning with text-based prompts and multi-modal conditioning. However, these works make one or more unnatural assumptions on the number and/or type of modality inputs used to express controllability. We propose InstructAny2Pix, a flexible multi-modal instruction-following system that enables users to edit an input image using instructions involving audio, images, and text. InstructAny2Pix consists of three building blocks that facilitate this capability: a multi-modal encoder that encodes different modalities such as images and audio into a unified latent space, a diffusion model that learns to decode representations in this latent space into images, and a multi-modal LLM that can understand instructions involving multiple images and audio pieces and generate a conditional embedding of the desired output, which can be used by the diffusion decoder. Additionally, to facilitate training efficiency and improve generation quality, we include an additional refinement prior module that enhances the visual quality of LLM outputs. These designs are critical to the performance of our system. We demonstrate that our system can perform a series of novel instruction-guided editing tasks. The code is available at https://github.com/jacklishufan/InstructAny2Pix.git
Generative Large Language Models (LLMs) have become the mainstream choice for fewshot and zeroshot learning thanks to the universality of text generation. Many users, however, do not need the broad capabilities of generative LLMs when they only want to automate a classification task. Smaller BERT-like models can also learn universal tasks, which allow them to do any text classification task without requiring fine-tuning (zeroshot classification) or to learn new tasks with only a few examples (fewshot), while being significantly more efficient than generative LLMs. This paper (1) explains how Natural Language Inference (NLI) can be used as a universal classification task that follows similar principles as instruction fine-tuning of generative LLMs, (2) provides a step-by-step guide with reusable Jupyter notebooks for building a universal classifier, and (3) shares the resulting universal classifier that is trained on 33 datasets with 389 diverse classes. Parts of the code we share has been used to train our older zeroshot classifiers that have been downloaded more than 55 million times via the Hugging Face Hub as of December 2023. Our new classifier improves zeroshot performance by 9.4%.
Accurate image classification and retrieval are of importance for clinical diagnosis and treatment decision-making. The recent contrastive language-image pretraining (CLIP) model has shown remarkable proficiency in understanding natural images. Drawing inspiration from CLIP, PathCLIP is specifically designed for pathology image analysis, utilizing over 200,000 image and text pairs in training. While the performance the PathCLIP is impressive, its robustness under a wide range of image corruptions remains unknown. Therefore, we conduct an extensive evaluation to analyze the performance of PathCLIP on various corrupted images from the datasets of Osteosarcoma and WSSS4LUAD. In our experiments, we introduce seven corruption types including brightness, contrast, Gaussian blur, resolution, saturation, hue, and markup at four severity levels. Through experiments, we find that PathCLIP is relatively robustness to image corruptions and surpasses OpenAI-CLIP and PLIP in zero-shot classification. Among the seven corruptions, blur and resolution can cause server performance degradation of the PathCLIP. This indicates that ensuring the quality of images is crucial before conducting a clinical test. Additionally, we assess the robustness of PathCLIP in the task of image-image retrieval, revealing that PathCLIP performs less effectively than PLIP on Osteosarcoma but performs better on WSSS4LUAD under diverse corruptions. Overall, PathCLIP presents impressive zero-shot classification and retrieval performance for pathology images, but appropriate care needs to be taken when using it. We hope this study provides a qualitative impression of PathCLIP and helps understand its differences from other CLIP models.