Recent advancements in large language models have showcased their remarkable generalizability across various domains. However, their reasoning abilities still have significant room for improvement, especially when confronted with scenarios requiring multi-step reasoning. Although large language models possess extensive knowledge, their reasoning often fails to effectively utilize this knowledge to establish a coherent thinking paradigm. These models sometimes show hallucinations as their reasoning procedures are unconstrained by logical principles. Aiming at improving the zero-shot chain-of-thought reasoning ability of large language models, we propose LoT (Logical Thoughts), a self-improvement prompting framework that leverages principles rooted in symbolic logic, particularly Reductio ad Absurdum, to systematically verify and rectify the reasoning processes step by step. Experimental evaluations conducted on language tasks in diverse domains, including arithmetic, commonsense, symbolic, causal inference, and social problems, demonstrate the efficacy of enhanced reasoning by logic. The implementation code for LoT can be accessed at: https://github.com/xf-zhao/LoT.

We study a class of reinforcement learning problems where the reward signals for policy learning are generated by a discriminator that is dependent on and jointly optimized with the policy. This interdependence between the policy and the discriminator leads to an unstable learning process because reward signals from an immature discriminator are noisy and impede policy learning, and conversely, an under-optimized policy impedes discriminator learning. We call this learning setting $\textit{Internally Rewarded Reinforcement Learning}$ (IRRL) as the reward is not provided directly by the environment but $\textit{internally}$ by the discriminator. In this paper, we formally formulate IRRL and present a class of problems that belong to IRRL. We theoretically derive and empirically analyze the effect of the reward function in IRRL and based on these analyses propose the clipped linear reward function. Experimental results show that the proposed reward function can consistently stabilize the training process by reducing the impact of reward noise, which leads to faster convergence and higher performance compared with baselines in diverse tasks.

Understanding spatial relations is essential for intelligent agents to act and communicate in the physical world. Relative directions are spatial relations that describe the relative positions of target objects with regard to the intrinsic orientation of reference objects. Grounding relative directions is more difficult than grounding absolute directions because it not only requires a model to detect objects in the image and to identify spatial relation based on this information, but it also needs to recognize the orientation of objects and integrate this information into the reasoning process. We investigate the challenging problem of grounding relative directions with end-to-end neural networks. To this end, we provide GRiD-3D, a novel dataset that features relative directions and complements existing visual question answering (VQA) datasets, such as CLEVR, that involve only absolute directions. We also provide baselines for the dataset with two established end-to-end VQA models. Experimental evaluations show that answering questions on relative directions is feasible when questions in the dataset simulate the necessary subtasks for grounding relative directions. We discover that those subtasks are learned in an order that reflects the steps of an intuitive pipeline for processing relative directions.

Neural networks can be powerful function approximators, which are able to model high-dimensional feature distributions from a subset of examples drawn from the target distribution. Naturally, they perform well at generalizing within the limits of their target function, but they often fail to generalize outside of the explicitly learned feature space. It is therefore an open research topic whether and how neural network-based architectures can be deployed for systematic reasoning. Many studies have shown evidence for poor generalization, but they often work with abstract data or are limited to single-channel input. Humans, however, learn and interact through a combination of multiple sensory modalities, and rarely rely on just one. To investigate compositional generalization in a multimodal setting, we generate an extensible dataset with multimodal input sequences from simulation. We investigate the influence of the underlying training data distribution on compostional generalization in a minimal LSTM-based network trained in a supervised, time continuous setting. We find compositional generalization to fail in simple setups while improving with the number of objects, actions, and particularly with a lot of color overlaps between objects. Furthermore, multimodality strongly improves compositional generalization in settings where a pure vision model struggles to generalize.

Reasoning about potential occlusions is essential for robots to efficiently predict whether an object exists in an environment. Though existing work shows that a robot with active perception can achieve various tasks, it is still unclear if occlusion reasoning can be achieved. To answer this question, we introduce the task of robotic object existence prediction: when being asked about an object, a robot needs to move as few steps as possible around a table with randomly placed objects to predict whether the queried object exists. To address this problem, we propose a novel recurrent neural network model that can be jointly trained with supervised and reinforcement learning methods using a curriculum training strategy. Experimental results show that 1) both active perception and occlusion reasoning are necessary to successfully achieve the task; 2) the proposed model demonstrates a good occlusion reasoning ability by achieving a similar prediction accuracy to an exhaustive exploration baseline while requiring only about 10% of the baseline’s number of movement steps on average; and 3) the model generalizes to novel object combinations with a moderate loss of accuracy.

One of the most remarkable findings in the literature on sentence embeddings has been that simple word vector averaging can compete with state-of-the-art models in many tasks. While counter-intuitive, a convincing explanation has been provided by Arora et al., who showed that the bag-of-words representation of a sentence can be recovered from its word vector average with almost perfect accuracy. Beyond word vector averaging, however, most sentence embedding models are essentially black boxes: while there is abundant empirical evidence about their strengths and weaknesses, it is not clear why and how different embedding strategies are able to capture particular properties of sentences. In this paper, we focus in particular on how sentence embedding models are able to capture word order. For instance, it seems intuitively puzzling that simple LSTM autoencoders are able to learn sentence vectors from which the original sentence can be reconstructed almost perfectly. With the aim of elucidating this phenomenon, we show that to capture word order, it is in fact sufficient to supplement standard word vector averages with averages of bigram and trigram vectors. To this end, we first study the problem of reconstructing bags-of-bigrams, focusing in particular on how suitable bigram vectors should be encoded. We then show that LSTMs are capable, in principle, of learning our proposed sentence embeddings. Empirically, we find that our embeddings outperform those learned by LSTM autoencoders on the task of sentence reconstruction, while needing almost no training data.

A key problem in computational sustainability is to understand the distribution of species across landscapes over time. This question gives rise to challenging large-scale prediction problems since (i) hundreds of species have to be simultaneously modeled and (ii) the survey data are usually inflated with zeros due to the absence of species for a large number of sites. The problem of tackling both issues simultaneously, which we refer to as the zero-inflated multi-target regression problem, has not been addressed by previous methods in statistics and machine learning. In this paper, we propose a novel deep model for the zero-inflated multi-target regression problem. To this end, we first model the joint distribution of multiple response variables as a multivariate probit model and then couple the positive outcomes with a multivariate log-normal distribution. By penalizing the difference between the two distributions’ covariance matrices, a link between both distributions is established. The whole model is cast as an end-to-end learning framework and we provide an efficient learning algorithm for our model that can be fully implemented on GPUs. We show that our model outperforms the existing state-of-the-art baselines on two challenging real-world species distribution datasets concerning bird and fish populations.

The Simple Temporal Problem (STP) is a fundamental temporal reasoning problem and has recently been extended to the Multiagent Simple Temporal Problem (MaSTP). In this paper we present a novel approach that is based on enforcing arc-consistency (AC) on the input (multiagent) simple temporal network. We show that the AC-based approach is sufficient for solving both the STP and MaSTP and provide efficient algorithms for them. As our AC-based approach does not impose new constraints between agents, it does not violate the privacy of the agents and is superior to the state-of-the-art approach to MaSTP. Empirical evaluations on diverse benchmark datasets also show that our AC-based algorithms for STP and MaSTP are significantly more efficient than existing approaches.

The capability of determining the right sequence of physical actions to achieve a given task is essential for AI that interacts with the physical world. The great difficulty in developing this capability has two main causes: (1) the world is continuous and therefore the action space is infinite, (2) due to noisy perception, we do not know the exact physical properties of our environment and therefore cannot precisely simulate the consequences of a physical action.

The Simple Temporal Problem (STP) has been widely used in various applications to schedule tasks. For dynamical systems, scheduling needs to be efficient and flexible to handle uncertainty and perturbation. To this end, modern approaches usually encode the temporal information as an STP instance. This representation contains redundant information, which can not only take a significant amount of storage space, but also make scheduling inefficient due to the non-concise representation. In this paper, we investigate the problem of simplifying an STP instance by removing redundant information. We show that such a simplification can result in a unique minimal representation without loss of temporal information, and present an efficient algorithm to achieve this task. Evaluation on a large benchmark dataset of STP exhibits a significant reduction in redundant information for the involved instances.

Whether reasoning with relative directions can be performed in NP has been an open problem in qualitative spatial reasoning. Efficient reasoning with relative directions is essential, for example, in rule-compliant agent navigation. In this paper, we prove that reasoning with relative directions is ∃R-complete. As a consequence, reasoning with relative directions is not in NP, unless NP=∃R.