Large Language Models (LLMs) exhibit strong performance across various natural language processing (NLP) tasks but remain vulnerable to hallucinations, generating factually incorrect or misleading outputs. Uncertainty estimation, often using predictive entropy estimation, is key to addressing this issue. However, existing methods often require multiple samples or extra computation to assess semantic entropy. This paper proposes an efficient, training-free uncertainty estimation method that approximates predictive entropy using the responses' top-K probabilities. Moreover, we employ an adaptive mechanism to determine K to enhance flexibility and filter out low-confidence probabilities. Experimental results on three free-form question-answering datasets across several LLMs demonstrate that our method outperforms expensive state-of-the-art baselines, contributing to the broader goal of enhancing LLM trustworthiness.

How can Large Language Models (LLMs) be aligned with human intentions and values? A typical solution is to gather human preference on model outputs and finetune the LLMs accordingly while ensuring that updates do not deviate too far from a reference model. Recent approaches, such as direct preference optimization (DPO), have eliminated the need for unstable and sluggish reinforcement learning optimization by introducing close-formed supervised losses. However, a significant limitation of the current approach is its design for a single reference model only, neglecting to leverage the collective power of numerous pretrained LLMs. To overcome this limitation, we introduce a novel closed-form formulation for direct preference optimization using multiple reference models. The resulting algorithm, Multi-Reference Preference Optimization (MRPO), leverages broader prior knowledge from diverse reference models, substantially enhancing preference learning capabilities compared to the single-reference DPO. Our experiments demonstrate that LLMs finetuned with MRPO generalize better in various preference data, regardless of data scarcity or abundance. Furthermore, MRPO effectively finetunes LLMs to exhibit superior performance in several downstream natural language processing tasks such as HH-RLHF, GSM8K and TruthfulQA.

Cross-domain offline reinforcement learning (RL) seeks to enhance sample efficiency in offline RL by utilizing additional offline source datasets. A key challenge is to identify and utilize source samples that are most relevant to the target domain. Existing approaches address this challenge by measuring domain gaps through domain classifiers, target transition dynamics modeling, or mutual information estimation using contrastive loss. However, these methods often require large target datasets, which is impractical in many real-world scenarios. In this work, we address cross-domain offline RL under a limited target data setting, identifying two primary challenges: (1) Dataset imbalance, which is caused by large source and small target datasets and leads to overfitting in neural network-based domain gap estimators, resulting in uninformative measurements; and (2) Partial domain overlap, where only a subset of the source data is closely aligned with the target domain. To overcome these issues, we propose DmC, a novel framework for cross-domain offline RL with limited target samples. Specifically, DmC utilizes k-nearest neighbor (k-NN) based estimation to measure domain proximity without neural network training, effectively mitigating overfitting. Then, by utilizing this domain proximity, we introduce a nearest-neighbor-guided diffusion model to generate additional source samples that are better aligned with the target domain, thus enhancing policy learning with more effective source samples. Through theoretical analysis and extensive experiments in diverse MuJoCo environments, we demonstrate that DmC significantly outperforms state-of-the-art cross-domain offline RL methods, achieving substantial performance gains.

Efficient molecular dynamics (MD) simulation is vital for understanding atomic-scale processes in materials science and biophysics. Traditional density functional theory (DFT) methods are computationally expensive, which limits the feasibility of long-term simulations. We propose a novel approach that formulates MD simulation as a time-series forecasting problem, enabling advanced forecasting models to predict atomic trajectories via displacements rather than absolute positions. We incorporate a physics-informed loss and inference mechanism based on DFT-parametrised pair-wise Morse potential functions that penalize unphysical atomic proximity to enforce physical plausibility. Our method consistently surpasses standard baselines in simulation accuracy across diverse materials. The results highlight the importance of incorporating physics knowledge to enhance the reliability and precision of atomic trajectory forecasting. Remarkably, it enables stable modeling of thousands of MD steps in minutes, offering a scalable alternative to costly DFT simulations.

Federated domain generalization (FedDG) addresses distribution shifts among clients in a federated learning frame-work. FedDG methods aggregate the parameters of locally trained client models to form a global model that generalizes to unseen clients while preserving data privacy. While improving the generalization capability of the global model, many existing approaches in FedDG jeopardize privacy by sharing statistics of client data between themselves. Our solution addresses this problem by contributing new ways to perform local client training and model aggregation. To improve local client training, we enforce (domain) invariance across local models with the help of a novel technique, latent space inversion, which enables better client privacy. When clients are not i.i.d, aggregating their local models may discard certain local adaptations. To overcome this, we propose an important weight aggregation strategy to prioritize parameters that significantly influence predictions of local models during aggregation. Our extensive experiments show that our approach achieves superior results over state-of-the-art methods with less communication overhead. Our code is available here.

Effective decision-making in partially observable environments demands robust memory management. Despite their success in supervised learning, current deep-learning memory models struggle in reinforcement learning environments that are partially observable and long-term. They fail to efficiently capture relevant past information, adapt flexibly to changing observations, and maintain stable updates over long episodes. We theoretically analyze the limitations of existing memory models within a unified framework and introduce the Stable Hadamard Memory, a novel memory model for reinforcement learning agents. Our model dynamically adjusts memory by erasing no longer needed experiences and reinforcing crucial ones computationally efficiently. To this end, we leverage the Hadamard product for calibrating and updating memory, specifically designed to enhance memory capacity while mitigating numerical and learning challenges. Our approach significantly outperforms state-of-the-art memory-based methods on challenging partially observable benchmarks, such as meta-reinforcement learning, long-horizon credit assignment, and POPGym, demonstrating superior performance in handling long-term and evolving contexts.

While Large Language Models (LLMs) excel at in-context learning (ICL) using just a few demonstrations, their performances are sensitive to demonstration orders. The reasons behind this sensitivity remain poorly understood. In this paper, we investigate the prompt embedding space to bridge the gap between the order sensitivity of ICL with inner workings of decoder-only LLMs, uncovering the clustering property: prompts sharing the first and last demonstrations have closer embeddings, with first-demonstration clustering usually being stronger in practice. We explain this property through extensive theoretical analyses and empirical evidences. Our finding suggests that the positional encoding and the causal attention mask are key contributors to the clustering phenomenon. Leveraging this clustering insight, we introduce Cluster-based Search, a novel method that accelerates the selection and ordering of demonstrations in self-adaptive ICL settings. Our approach substantially decreases the time complexity from factorial to quadratic, saving 92% to nearly 100% execution time while maintaining comparable performance to exhaustive search.

Artificial agents with the aid of large language models (LLMs) are effective in various real-world scenarios but struggle to cooperate in social dilemmas. When making decisions under the strain of selecting between long-term consequences and short-term benefits in commonly shared resources, LLM-based agents often exploit the environment, leading to early depletion. Inspired by the concept of consideration of future consequences (CFC), which is well-known in social psychology, we propose a framework to enable the ability to consider future consequences for LLM-based agents, which results in a new kind of agent that we term the CFC-Agent. We enable the CFC-Agent to act toward different levels of consideration for future consequences. Our first set of experiments, where LLM is directly asked to make decisions, shows that agents considering future consequences exhibit sustainable behaviour and achieve high common rewards for the population. Extensive experiments in complex environments showed that the CFC-Agent can manage a sequence of calls to LLM for reasoning and engaging in communication to cooperate with others to resolve the common dilemma better. Finally, our analysis showed that considering future consequences not only affects the final decision but also improves the conversations between LLM-based agents toward a better resolution of social dilemmas.

Decision Transformers (DT) play a crucial role in modern reinforcement learning, leveraging offline datasets to achieve impressive results across various domains. However, DT requires high-quality, comprehensive data to perform optimally. In real-world applications, the lack of training data and the scarcity of optimal behaviours make training on offline datasets challenging, as suboptimal data can hinder performance. To address this, we propose the Counterfactual Reasoning Decision Transformer (CRDT), a novel framework inspired by counterfactual reasoning. CRDT enhances DT’s ability to reason beyond known data by generating and utilizing counterfactual experiences, enabling improved decision-making in unseen scenarios. Experiments across Atari and D4RL benchmarks, including scenarios with limited data and altered dynamics, demonstrate that CRDT outperforms conventional DT approaches. Additionally, reasoning counterfactually allows the DT agent to obtain stitching abilities, combining suboptimal trajectories, without architectural modifications. These results highlight the potential of counterfactual reasoning to enhance reinforcement learning agents' performance and generalization capabilities.

Fair Domain Generalization with Heterogeneous Sensitive Attributes Across Domains

Abstract Large language models (LLMs) have recently demonstrated their impressive ability to provide context-aware responses via text. This ability could potentially be used to predict plausible solutions in sequential decision making tasks pertaining to pattern completion. For example, by observing a partial stack of cubes, LLMs can predict the correct sequence in which the remaining cubes should be stacked by extrapolating the observed patterns (e.g., cube sizes, colors or other attributes) in the partial stack. In this work, we introduce LaGR (language-guided reinforcement learning), which uses this predictive ability of LLMs to propose solutions to tasks that have been partially completed by a primary reinforcement learning (RL) agent, in order to subsequently guide the latter’s training. However, as RL training is generally not sample-efficient, deploying this approach would inherently imply that the LLM be repeatedly queried for solutions; a process that can be expensive and infeasible. To address this issue, we introduce SEQ (sample-efficient querying), where we simultaneously train a secondary RL agent to decide when the LLM should be queried for solutions. Specifically, we use the quality of the solutions emanating from the LLM as the reward to train this agent. We show that our proposed framework LaGR-SEQ enables more efficient primary RL training, while simultaneously minimizing the number of queries to the LLM. We demonstrate our approach on a series of tasks and highlight the advantages of our approach, along with its limitations and potential future research directions.

In this paper, we study the “dataset bias” problem from a statistical standpoint, and identify the main cause of the problem as the strong correlation between a class attribute u and a non-class attribute b in the input x, represented by p(u|b) differing significantly from p(u). Since p(u|b) appears as part of the sampling distributions in the standard maximum log-likelihood (MLL) objective, a model trained on a biased dataset via MLL inherently incorporates such correlation into its parameters, leading to poor generalization to unbiased test data. From this observation, we propose to mitigate dataset bias via either weighting the objective of each sample n by 1 / p(un|bn) or sampling that sample with a weight proportional to 1 / p(un|bn). While both methods are statistically equivalent, the former proves more stable and effective in practice. Additionally, we establish a connection between our debiasing approach and causal reasoning, reinforcing our method’s theoretical foundation. However, when the bias label is unavailable, computing p(u|b) exactly is difficult. To overcome this challenge, we propose to approximate 1 / p(u|b) using a biased classifier trained with “bias amplification” losses. Extensive experiments on various biased datasets demonstrate the superiority of our method over existing debiasing techniques in most settings, validating our theoretical analysis.

Prompt optimization is essential for enhancing the performance of Large Language Models (LLMs) in a range of Natural Language Processing (NLP) tasks, particularly in scenarios of few-shot learning where training examples are incorporated directly into the prompt. Despite the growing interest in optimizing prompts with few-shot examples, existing methods for prompt optimization are often resource-intensive or perform inadequately. In this work, we propose PrOmpting with Episodic Memory (POEM), a novel prompt optimization technique that is simple, efficient, and demonstrates strong generalization capabilities. We approach prompt optimization as a Reinforcement Learning (RL) challenge, using episodic memory to archive combinations of input data, permutations of few-shot examples, and the rewards observed during training. In the testing phase, we optimize the sequence of examples for each test query by selecting the sequence that yields the highest total rewards from the top-k most similar training examples in the episodic memory. Our results show that POEM outperforms recent techniques like TEMPERA and RLPrompt by over 5.3% in various text classification tasks. Furthermore, our approach adapts well to broader language understanding tasks, consistently outperforming conventional heuristic methods for ordering examples.

Businesses often need to query visually rich documents (VRDs), e.g., purchase receipts, medical records, and insurance forms, among many other forms from multiple vendors, to make informed decisions. As such, several techniques have been proposed to automatically extract independent entities of interest from VRDs such as extracting price tags from purchase receipts, etc. However, for extracting semantically linked entities, such as finding corresponding price tags for each item, these techniques either have limited capability in handling new layouts, e.g., template-based approaches, or require extensive amounts of pre-training data and do not perform well, e.g., deep-learning approaches. In this work, we introduce a program synthesis method, namely VRDSynth, to automatically generate programs to extract entity relations from multilingual VRDs. Two key novelties, which empower VRDSynth to tackle flexible layouts while requiring no pre-training data for extracting entity relations, include: (1) a new domain-specific language (DSL) to effectively capture the spatial and textual relations between document entities, and (2) a novel synthesis algorithm that makes use of frequent spatial relations between entities to construct initial programs, equivalent reduction to prune the search space, and a combination of positive, negative, and mutually exclusive programs to improve the coverage of programs. We evaluate our method on two popular VRD understanding benchmarks, namely FUNSD and XFUND, on the semantic entity linking task, consisting of 1,600 forms in 8 different languages. Experiments show that VRDSynth, despite having no prior pre-training data, outperforms the state-of-the-art pre-trained deep-learning approach, namely LayoutXLM, in 5 out of 8 languages. Noticeably, VRDSynth achieved an improvement of 42% over LayoutXLM in terms of F1 score on FUNSD while being complementary to LayoutXLM in 7/8 languages. Regarding efficiency, VRDSynth significantly improves the memory footprint required for storage and inference over LayoutXLM (1M and 380MB versus that of 1.48GB and 3GB required by LayoutXLM), while maintaining similar time efficiency despite the speed differences between the languages used for implementation (Python vs C++).

Social reasoning necessitates the capacity of theory of mind (ToM), the ability to contextualise and attribute mental states to others without having access to their internal cognitive structure. Recent machine learning approaches to ToM have demonstrated that we can train the observer to read the past and present behaviours of other agents and infer their beliefs (including false beliefs about things that no longer exist), goals, intentions and future actions. The challenges arise when the behavioural space is complex, demanding skilful space navigation for rapidly changing contexts for an extended period. We tackle the challenges by equipping the observer with novel neural memory mechanisms to encode, and hierarchical attention to selectively retrieve information about others. The memories allow rapid, selective querying of distal related past behaviours of others to deliberatively reason about their current mental state, beliefs and future behaviours. This results in ToMMY, a theory of mind model that learns to reason while making little assumptions about the underlying mental processes. We also construct a new suite of experiments to demonstrate that memories facilitate the learning process and achieve better theory of mind performance, especially for high-demand false-belief tasks that require inferring through multiple steps of changes.

Understanding other agents is a key challenge in constructing artificial social agents. Current works focus on centralised training, wherein agents are allowed to know all the information about others and the environmental state during training. In contrast, this work studies decentralised training, wherein agents must learn the model of other agents in order to cooperate with them under partially-observable conditions, even during training, i.e. learning agents are myopic. The intrinsic motivation for artificial agents is modelled on the concept of human social motivation that entices humans to meet and understand each other, especially when experiencing a utility loss. Our intrinsic motivation encourages agents to stay near each other to obtain better observations and construct a model of others. They do so when their model of other agents is poor, or the overall task performance is bad during the learning phase. This simple but effective method facilitates the processes of modelling others, resulting in an improvement of the performance in cooperative tasks significantly. Our experiments demonstrate that the socially-motivated agent can model others better and promote cooperation across different tasks.

The optimistic nature of the Q−learning target leads to an overestimation bias, which is an inherent problem associated with standard Q−learning. Such a bias fails to account for the possibility of low returns, particularly in risky scenarios. However, the existence of biases, whether overestimation or underestimation, need not necessarily be undesirable. In this paper, we analytically examine the utility of biased learning, and show that specific types of biases may be preferable, depending on the scenario. Based on this finding, we design a novel reinforcement learning algorithm, Balanced Q-learning, in which the target is modified to be a convex combination of a pessimistic and an optimistic term, whose associated weights are determined online, analytically. Such a balanced target inherently promotes risk-averse behavior, which we examine through the lens of the agent's exploration. We prove the convergence of this algorithm in a tabular setting, and empirically demonstrate its consistently good learning performance in various environments.

We introduce a novel training procedure for policy gradient methods wherein episodic memory is used to optimize the hyperparameters of reinforcement learning algorithms on-the-fly. Unlike other hyperparameter searches, we formulate hyperparameter scheduling as a standard Markov Decision Process and use episodic memory to store the outcome of used hyperparameters and their training contexts. At any policy update step, the policy learner refers to the stored experiences, and adaptively reconfigures its learning algorithm with the new hyperparameters determined by the memory. This mechanism, dubbed as Episodic Policy Gradient Training (EPGT), enables an episodic learning process, and jointly learns the policy and the learning algorithm's hyperparameters within a single run. Experimental results on both continuous and discrete environments demonstrate the advantage of using the proposed method in boosting the performance of various policy gradient algorithms.

Artificial Neural Networks are functionally equivalent to special-purpose computers. Their inter-neuronal connection weights represent the learnt Neural Program that instructs the networks on how to compute the data. However, without storing Neural Programs, they are restricted to only one, overwriting learnt programs when trained on new data. Here we design Neurocoder, a new class of general-purpose neural networks in which the neural network “codes” itself in a data-responsive way by composing relevant programs from a set of shareable, modular programs stored in external memory. This time, a Neural Program is efficiently treated as data in memory. Integrating Neurocoder into current neural architectures, we demonstrate new capacity to learn modular programs, reuse simple programs to build complex ones, handle pattern shifts and remember old programs as new ones are learnt, and show substantial performance improvement in solving object recognition, playing video games and continual learning tasks.

Data-free Knowledge Distillation (DFKD) has attracted attention recently thanks to its appealing capability of transferring knowledge from a teacher network to a student network without using training data. The main idea is to use a generator to synthesize data for training the student. As the generator gets updated, the distribution of synthetic data will change. Such distribution shift could be large if the generator and the student are trained adversarially, causing the student to forget the knowledge it acquired at the previous steps. To alleviate this problem, we propose a simple yet effective method called Momentum Adversarial Distillation (MAD) which maintains an exponential moving average (EMA) copy of the generator and uses synthetic samples from both the generator and the EMA generator to train the student. Since the EMA generator can be considered as an ensemble of the generator's old versions and often undergoes a smaller change in updates compared to the generator, training on its synthetic samples can help the student recall the past knowledge and prevent the student from adapting too quickly to the new updates of the generator. Our experiments on six benchmark datasets including big datasets like ImageNet and Places365 demonstrate the superior performance of MAD over competing methods for handling the large distribution shift problem. Our method also compares favorably to existing DFKD methods and even achieves state-of-the-art results in some cases.

We introduce a constrained optimization method for policy gradient reinforcement learning, which uses two trust regions to regulate each policy update. In addition to using the proximity of one single old policy as the first trust region as done by prior works, we propose forming a second trust region by constructing another virtual policy that represents a wide range of past policies. We then enforce the new policy to stay closer to the virtual policy, which is beneficial if the old policy performs poorly. We propose a mechanism to automatically build the virtual policy from a memory buffer of past policies, providing a new capability for dynamically selecting appropriate trust regions during the optimization process. Our proposed method, dubbed Memory-Constrained Policy Optimization (MCPO), is examined in diverse environments, including robotic locomotion control, navigation with sparse rewards and Atari games, consistently demonstrating competitive performance against recent on-policy constrained policy gradient methods.

The capacity to achieve out-of-distribution (OOD) generalization is a hallmark of human intelligence and yet remains out of reach for machines. This remarkable capability has been attributed to our abilities to make conceptual abstraction and analogy, and to a mechanism known as indirection, which binds two representations and uses one representation to refer to the other. Inspired by these mechanisms, we hypothesize that OOD generalization may be achieved by performing analogy-making and indirection in the functional space instead of the data space as in current methods. To realize this, we design FINE (Functional Indirection Neural Estimator), a neural framework that learns to compose functions that map data input to output on-the-fly. FINE consists of a backbone network and a trainable semantic memory of basis weight matrices. Upon seeing a new input-output data pair, FINE dynamically constructs the backbone weights by mixing the basis weights. The mixing coefficients are indirectly computed through querying a separate corresponding semantic memory using the data pair. We demonstrate empirically that FINE can strongly improve out-of-distribution generalization on IQ tasks that involve geometric transformations. In particular, we train FINE and competing models on IQ tasks using images from the MNIST, Omniglot and CIFAR100 datasets and test on tasks with unseen image classes from one or different datasets and unseen transformation rules. FINE not only achieves the best performance on all tasks but also is able to adapt to small-scale data scenarios.

Transfer in reinforcement learning is usually achieved through generalisation across tasks. Whilst many studies have investigated transferring knowledge when the reward function changes, they have assumed that the dynamics of the environments remain consistent. Many real-world RL problems require transfer among environments with different dynamics. To address this problem, we propose an approach based on successor features in which we model successor feature functions with Gaussian Processes permitting the source successor features to be treated as noisy measurements of the target successor feature function. Our theoretical analysis proves the convergence of this approach as well as the bounded error on modelling successor feature functions with Gaussian Processes in environments with both different dynamics and rewards. We demonstrate our method on benchmark datasets and show that it outperforms current baselines.

The rise of big data and big compute has brought modern neural networks to many walks of digital life, thanks to the relative ease of constructing large models that scale to the real world. Current successes of Transformers and self-supervised pretraining on massive data have led some to believe that deep neural networks will be able to do almost everything once we have sufficient data and computational resources. However, neural networks are fast to exploit surface statistics but fail miserably to generalize to novel combinations. This is because they are not designed for deliberate reasoning -- the capacity to deliberately deduce new knowledge out of the contextualized data. This tutorial reviews recent developments to extend the capacity of neural networks to "learning-to-reason'' from data, where the task is to determine if the data entails a conclusion. This capacity opens up new ways to generate insights from data through arbitrary compositional querying without the need of predefining a narrow set of tasks. The tutorial consists of four parts. The first part covers the learning-to-reason framework, and explains how neural networks can serve as a strong backbone for reasoning through its natural operations such as binding, attention & dynamic computational graphs. The second part goes into more detail on how neural networks perform reasoning over unstructured and structured data, and across modalities. The third part reviews neural memories and their role in reasoning. The last part discusses generalization to novel combinations, under less supervision and with more knowledge.

Episodic control enables sample efficiency in reinforcement learning by recalling past experiences from an episodic memory. We propose a new model-based episodic memory of trajectories addressing current limitations of episodic control. Our memory estimates trajectory values, guiding the agent towards good policies. Built upon the memory, we construct a complementary learning model via a dynamic hybrid control unifying model-based, episodic and habitual learning into a single architecture. Experiments demonstrate that our model allows significantly faster and better learning than other strong reinforcement learning agents across a variety of environments including stochastic and non-Markovian settings.

Heretofore, neural networks with external memory are restricted to single memory with lossy representations of memory interactions. A rich representation of relationships between memory pieces urges a high-order and segregated relational memory. In this paper, we propose to separate the storage of individual experiences (item memory) and their occurring relationships (relational memory). The idea is implemented through a novel Self-attentive Associative Memory (SAM) operator. Found upon outer product, SAM forms a set of associative memories that represent the hypothetical high-order relationships between arbitrary pairs of memory elements, through which a relational memory is constructed from an item memory. The two memories are wired into a single sequential model capable of both memorization and relational reasoning. We achieve competitive results with our proposed two-memory model in a diversity of machine learning tasks, from challenging synthetic problems to practical testbeds such as geometry, graph, reinforcement learning, and question answering.

One of the biggest obstacles in Chinese and Japanese text line recognition is how to present their enormous character sets. The most common solution is to merely choose and represent a small subset of characters using one-hot encoding. However, such an approach is costly to describe huge character sets, and ignores their semantic relationships. Recent studies have attempted to utilize different encoding methods, but they struggle to build a bijection mapping. In this work, we propose a novel encoding method, called LOgographic DEComposition encoding (LODEC), that can efficiently perform a 1-to-1 mapping for all Chinese and Japanese characters. As such, LODEC enables to encode over 21,000 Chinese and Japanese characters by 520 fundamental elements. Moreover, to handle the vast style variety of handwritten texts in the two languages, we propose a novel deep learning (DL) architecture, called LODENet, together with an end-to-end training scheme, that leverages auxiliary ground truths generated by LODEC or other radical-based encoding methods. We systematically performed experiments on both Chinese and Japanese datasets, and found that our approach surpassed the performance of state-of-the-art baselines. Furthermore, empirical evidence shows that our method can gain significantly improvement using synthesized text line images without the need for domain knowledge.

One of the core tasks in multi-view learning is to capture relations among views. For sequential data, the relations not only span across views, but also extend throughout the view length to form long-term intra-view and inter-view interactions. In this paper, we present a new memory augmented neural network that aims to model these complex interactions between two asynchronous sequential views. Our model uses two encoders for reading from and writing to two external memories for encoding input views. The intra-view interactions and the long-term dependencies are captured by the use of memories during this encoding process. There are two modes of memory accessing in our system: late-fusion and early-fusion, corresponding to late and early inter-view interactions. In the late-fusion mode, the two memories are separated, containing only view-specific contents. In the early-fusion mode, the two memories share the same addressing space, allowing cross-memory accessing. In both cases, the knowledge from the memories will be combined by a decoder to make predictions over the output space. The resulting dual memory neural computer is demonstrated on a comprehensive set of experiments, including a synthetic task of summing two sequences and the tasks of drug prescription and disease progression in healthcare. The results demonstrate competitive performance over both traditional algorithms and deep learning methods designed for multi-view problems.

Introducing variability while maintaining coherence is a core task in learning to generate utterances in conversation. Standard neural encoder-decoder models and their extensions using conditional variational autoencoder often result in either trivial or digressive responses. To overcome this, we explore a novel approach that injects variability into neural encoder-decoder via the use of external memory as a mixture model, namely Variational Memory Encoder-Decoder (VMED). By associating each memory read with a mode in the latent mixture distribution at each timestep, our model can capture the variability observed in sequential data such as natural conversations. We empirically compare the proposed model against other recent approaches on various conversational datasets. The results show that VMED consistently achieves significant improvement over others in both metric-based and qualitative evaluations.