Recently, Large Language Models (LLMs) have achieved amazing zero-shot learning performance over a variety of Natural Language Processing (NLP) tasks, especially for text generative tasks. Yet, the large size of LLMs often leads to the high computational cost of model training and online deployment. In our work, we present ALTER, a system that effectively builds the multi-tAsk Learners with mixTure-of-task-adaptERs upon small language models (with <1B parameters) to address multiple NLP tasks simultaneously, capturing the commonalities and differences between tasks, in order to support domain-specific applications. Specifically, in ALTER, we propose the Mixture-of-Task-Adapters (MTA) module as an extension to the transformer architecture for the underlying model to capture the intra-task and inter-task knowledge. A two-stage training method is further proposed to optimize the collaboration between adapters at a small computational cost. Experimental results over a mixture of NLP tasks show that our proposed MTA architecture and the two-stage training method achieve good performance. Based on ALTER, we have also produced MTA-equipped language models for various domains.
The excellent text-to-image synthesis capability of diffusion models has driven progress in synthesizing coherent visual stories. The current state-of-the-art method combines the features of historical captions, historical frames, and the current captions as conditions for generating the current frame. However, this method treats each historical frame and caption as the same contribution. It connects them in order with equal weights, ignoring that not all historical conditions are associated with the generation of the current frame. To address this issue, we propose Causal-Story. This model incorporates a local causal attention mechanism that considers the causal relationship between previous captions, frames, and current captions. By assigning weights based on this relationship, Causal-Story generates the current frame, thereby improving the global consistency of story generation. We evaluated our model on the PororoSV and FlintstonesSV datasets and obtained state-of-the-art FID scores, and the generated frames also demonstrate better storytelling in visuals. The source code of Causal-Story can be obtained from https://github.com/styufo/Causal-Story.
Large Language Models (LLMs) have shown immense potential in multimodal applications, yet the convergence of textual and musical domains remains relatively unexplored. To address this gap, we present MusiLingo, a novel system for music caption generation and music-related query responses. MusiLingo employs a single projection layer to align music representations from the pre-trained frozen music audio model MERT with the frozen LLaMA language model, bridging the gap between music audio and textual contexts. We train it on an extensive music caption dataset and fine-tune it with instructional data. Due to the scarcity of high-quality music Q&A datasets, we created the MusicInstruct (MI) dataset from MusicCaps, tailored for open-ended music inquiries. Empirical evaluations demonstrate its competitive performance in generating music captions and composing music-related Q&A pairs. Our introduced dataset enables notable advancements beyond previous ones.
Individual differences in personality determine our preferences, traits and values, which should similarly hold for the way we express ourselves. With current advancements and transformations of technology and society, text-based communication has become ordinary and often even surpasses natural voice conversations -- with distinct challenges and opportunities. In this exploratory work, we investigate the impact of personality on the tendency how players of a team-based collaborative alternate reality game express themselves affectively. We collected chat logs from eleven players over two weeks, labeled them according to their affective state, and assessed the connection between them and the five-factor personality domains and facets. After applying multi-linear regression, we found a series of reasonable correlations between (combinations of) personality variables and expressed affect -- as increased confusion could be predicted by lower self-competence (C1), personal annoyance by vulnerability to stress (N6) and expressing anger occured more often in players that are prone to anxiety (N1), less humble and modest (A5), think less carefully before they act (C6) and have higher neuroticism (N). Expanding the data set, sample size and input modalities in subsequent work, we aim to confirm these findings and reveal even more interesting connections that could inform affective computing and games user research equally.
Emotion estimation in images is a challenging task, typically using computer vision methods to directly estimate people's emotions using face, body pose and contextual cues. In this paper, we explore whether Large Language Models (LLMs) can support the contextual emotion estimation task, by first captioning images, then using an LLM for inference. First, we must understand: how well do LLMs perceive human emotions? And which parts of the information enable them to determine emotions? One initial challenge is to construct a caption that describes a person within a scene with information relevant for emotion perception. Towards this goal, we propose a set of natural language descriptors for faces, bodies, interactions, and environments. We use them to manually generate captions and emotion annotations for a subset of 331 images from the EMOTIC dataset. These captions offer an interpretable representation for emotion estimation, towards understanding how elements of a scene affect emotion perception in LLMs and beyond. Secondly, we test the capability of a large language model to infer an emotion from the resulting image captions. We find that GPT-3.5, specifically the text-davinci-003 model, provides surprisingly reasonable emotion predictions consistent with human annotations, but accuracy can depend on the emotion concept. Overall, the results suggest promise in the image captioning and LLM approach.
3D visual grounding is a critical skill for household robots, enabling them to navigate, manipulate objects, and answer questions based on their environment. While existing approaches often rely on extensive labeled data or exhibit limitations in handling complex language queries, we propose LLM-Grounder, a novel zero-shot, open-vocabulary, Large Language Model (LLM)-based 3D visual grounding pipeline. LLM-Grounder utilizes an LLM to decompose complex natural language queries into semantic constituents and employs a visual grounding tool, such as OpenScene or LERF, to identify objects in a 3D scene. The LLM then evaluates the spatial and commonsense relations among the proposed objects to make a final grounding decision. Our method does not require any labeled training data and can generalize to novel 3D scenes and arbitrary text queries. We evaluate LLM-Grounder on the ScanRefer benchmark and demonstrate state-of-the-art zero-shot grounding accuracy. Our findings indicate that LLMs significantly improve the grounding capability, especially for complex language queries, making LLM-Grounder an effective approach for 3D vision-language tasks in robotics. Videos and interactive demos can be found on the project website https://chat-with-nerf.github.io/ .
Large Language Models (LLMs) present significant priority in text understanding and generation. However, LLMs suffer from the risk of generating harmful contents especially while being employed to applications. There are several black-box attack methods, such as Prompt Attack, which can change the behaviour of LLMs and induce LLMs to generate unexpected answers with harmful contents. Researchers are interested in Prompt Attack and Defense with LLMs, while there is no publicly available dataset to evaluate the abilities of defending prompt attack. In this paper, we introduce a Chinese Prompt Attack Dataset for LLMs, called CPAD. Our prompts aim to induce LLMs to generate unexpected outputs with several carefully designed prompt attack approaches and widely concerned attacking contents. Different from previous datasets involving safety estimation, We construct the prompts considering three dimensions: contents, attacking methods and goals, thus the responses can be easily evaluated and analysed. We run several well-known Chinese LLMs on our dataset, and the results show that our prompts are significantly harmful to LLMs, with around 70% attack success rate. We will release CPAD to encourage further studies on prompt attack and defense.
The rising demand for creating lifelike avatars in the digital realm has led to an increased need for generating high-quality human videos guided by textual descriptions and poses. We propose Dancing Avatar, designed to fabricate human motion videos driven by poses and textual cues. Our approach employs a pretrained T2I diffusion model to generate each video frame in an autoregressive fashion. The crux of innovation lies in our adept utilization of the T2I diffusion model for producing video frames successively while preserving contextual relevance. We surmount the hurdles posed by maintaining human character and clothing consistency across varying poses, along with upholding the background's continuity amidst diverse human movements. To ensure consistent human appearances across the entire video, we devise an intra-frame alignment module. This module assimilates text-guided synthesized human character knowledge into the pretrained T2I diffusion model, synergizing insights from ChatGPT. For preserving background continuity, we put forth a background alignment pipeline, amalgamating insights from segment anything and image inpainting techniques. Furthermore, we propose an inter-frame alignment module that draws inspiration from an auto-regressive pipeline to augment temporal consistency between adjacent frames, where the preceding frame guides the synthesis process of the current frame. Comparisons with state-of-the-art methods demonstrate that Dancing Avatar exhibits the capacity to generate human videos with markedly superior quality, both in terms of human and background fidelity, as well as temporal coherence compared to existing state-of-the-art approaches.
Domain adaptation using text-only corpus is challenging in end-to-end(E2E) speech recognition. Adaptation by synthesizing audio from text through TTS is resource-consuming. We present a method to learn Unified Speech-Text Representation in Conformer Transducer(USTR-CT) to enable fast domain adaptation using the text-only corpus. Different from the previous textogram method, an extra text encoder is introduced in our work to learn text representation and is removed during inference, so there is no modification for online deployment. To improve the efficiency of adaptation, single-step and multi-step adaptations are also explored. The experiments on adapting LibriSpeech to SPGISpeech show the proposed method reduces the word error rate(WER) by relatively 44% on the target domain, which is better than those of TTS method and textogram method. Also, it is shown the proposed method can be combined with internal language model estimation(ILME) to further improve the performance.
The emergence of human-like abilities of AI systems for content generation in domains such as text, audio, and vision has prompted the development of classifiers to determine whether content originated from a human or a machine. Implicit in these efforts is an assumption that the generation properties of a human are different from that of the machine. In this work, we provide a framework in the language of statistical pattern recognition that quantifies the difference between the distributions of human and machine-generated content conditioned on an evaluation context. We describe current methods in the context of the framework and demonstrate how to use the framework to evaluate the progression of generative models towards human-like capabilities, among many axes of analysis.