Combining logical reasoning with large language models enhances their capacity to solve complex problems reliably and robustly. Nevertheless, previous studies for logical reasoning mainly focus on capturing entity-aware knowledge while needing help to capture logical relationships among facts. Besides, the limited labeled data further hinders the ability of data-hungry neural models. Therefore, we propose a novel symbolic-neural reasoning method based on factual evidence and Abstract Meaning Representation (AMR) to enhance the Reasoning ability of Large Language Models. The former enriches the semantics of the context to obtain implicit logical expressions according to factual evidence and logical rules. The latter exploit AMR to augment literally similar but logically different training instances for contrastive learning. Extensive experiments on benchmark datasets demonstrate that the proposed method significantly outperforms solid baselines and achieves new state-of-the-art performance.

Although research has produced promising results demonstrating the utility of active inference (AIF) in Markov decision processes (MDPs), there is relatively less work that builds AIF models in the context of environments and problems that take the form of partially observable Markov decision processes (POMDPs). In POMDP scenarios, the agent must infer the unobserved environmental state from raw sensory observations, e.g., pixels in an image. Additionally, less work exists in examining the most difficult form of POMDP-centered control: continuous action space POMDPs under sparse reward signals. In this work, we address issues facing the AIF modeling paradigm by introducing novel prior preference learning techniques and self-revision schedules to help the agent excel in sparse-reward, continuous action, goal-based robotic control POMDP environments. Empirically, we show that our agents offer improved performance over state-of-the-art models in terms of cumulative rewards, relative stability, and success rate.

Abstractive Dialogue summarization is a challenging task due to its dynamic interaction among multiple speakers and lack of sufficient training data. Existing methods treat the dialogue as a linear sequence of utterances while ignoring the diverse topic relations between each utterance. Besides, the limited labeled data further hinders the ability of data hungry neural models. Therefore, we propose a novel Topic Paraphrasing Model (TPDSM) for abstractive dialogue summarization. It adopts cross attention to interactively acquire other topic critical information. Moreover, we construct a topic Paraphrasing corpus between topic and narrative sentence from existing datasets. Extensive experiments on benchmark datasets demonstrate that the proposed model significantly outperforms strong baselines and achieves new state-of-the-art performance.

Deep neural networks for video-based eye tracking have demonstrated resilience to noisy environments, stray reflections, and low resolution. However, to train these networks, a large number of manually annotated images are required. To alleviate the cumbersome process of manual labeling, computer graphics rendering is employed to automatically generate a large corpus of annotated eye images under various conditions. In this work, we introduce a synthetic eye image generation platform that improves upon previous work by adding features such as an active deformable iris, an aspherical cornea, retinal retro-reflection, gaze-coordinated eye-lid deformations, and blinks. To demonstrate the utility of our platform, we render images reflecting the represented gaze distributions inherent in two publicly available datasets, NVGaze and OpenEDS. We also report on the performance of two semantic segmentation architectures (SegNet and RITnet) trained on rendered images and tested on the original datasets.

Driven by significant investments from the gaming, film, advertising, and customer service industries among others, efforts across many different fields are converging to create realistic representations of humans that look like (computer graphics), sound like (natural language generation), move like (motion capture), and reason like (artificial intelligence) real humans. The ultimate goal of this work is to push the boundaries even further by exploring the development of realistic self-organized virtual humans that are capable of demonstrating coordinated behaviors across different modalities. Eye movements, for example, may be accompanied by changes in facial expression, head orientation, posture, gait properties, or speech. Traditionally however, these modalities are captured and modeled separately and this disconnect contributes to the well-known uncanny valley phenomenon. We focus initially on facial modalities, in particular, coordinated eye and head movements (and eventually facial expressions), but our proposed data-driven framework will be able to accommodate other modalities as well. transfer [Laine et al. 2017]. Despite these advances, the resulting renderings or animations are often still distinguishable from a real human, sometimes in unsettling ways - the so called uncanny valley phenomenon [Mori et al. 2012]. We argue that the traditional approach of capturing and modeling various human modalities separately contributes this effect. In this work, we focus on capturing, transferring, and generating realistic coordinated facial modalities (eye movements, head movements, and eventually facial expressions). We envision a flexible framework that can be extended to accommodate other modalities as well.