IJCAI 2023 Papers

Advancements in reinforcement learning (RL) have demonstrated superhuman performance in complex tasks such as Starcraft, Go, Chess etc. However, knowledge transfer from Artificial ``Experts" to humans remain a significant challenge. A promising avenue for such transfer would be the use of curricula. Recent methods in curricula generation focuses on training RL agents efficiently, yet such methods rely on surrogate measures to track student progress, and are not suited for training robots in the real world (or more ambitiously humans). In this paper, we introduce a method named Parameterized Environment Response Model (PERM) that shows promising results in training RL agents in parameterized environments. Inspired by Item Response Theory, PERM seeks to model difficulty of environments and ability of RL agents directly. Given that RL agents and humans are trained more efficiently under the ``zone of proximal development", our method generates a curriculum by matching the difficulty of an environment to the current ability of the student. In addition, PERM can be trained offline and does not employ non-stationary measures of student ability, making it suitable for transfer between students. We demonstrate PERM's ability to represent the environment parameter space, and training with RL agents with PERM produces a strong performance in deterministic environments. Lastly, we show that our method is transferable between students, without any sacrifice in training quality.

The tractability of the lightweight description logic EL has allowed for the construction of large and widely used ontologies that support semantic interoperability. However, comprehensive domains with a broad user base are often at odds with strong axiomatisations otherwise useful for inferencing, since these are usually context dependent and subject to diverging perspectives. In this paper we introduce Standpoint EL, a multi-modal extension of EL that allows for the integrated representation of domain knowledge relative to diverse, possibly conflicting standpoints (or contexts), which can be hierarchically organised and put in relation to each other. We establish that Standpoint EL still exhibits EL's favourable PTime standard reasoning, whereas introducing additional features like empty standpoints, rigid roles, and nominals makes standard reasoning tasks intractable.

Tracking individuals is a vital part of many experiments conducted to understand collective behaviour. Ants are the paradigmatic model system for such experiments but their lack of individually distinguishing visual features and their high colony densities make it extremely difficult to perform reliable racking automatically. Additionally, the wide diversity of their species' appearances makes a generalized approach even harder. In this paper, we propose a data-driven multi-object tracker that, for the first time, employs domain adaptation to achieve the required generalisation. This approach is built upon a joint-detection-and-tracking framework that is extended by a set of domain discriminator modules integrating an adversarial training strategy in addition to the tracking loss. In addition to this novel domain-adaptive tracking framework, we present a new dataset and a benchmark for the ant tracking problem. The dataset contains 57 video sequences with full trajectory annotation, including 30k frames captured from two different ant species moving on different background patterns. It comprises 33 and 24 sequences for source and target domains, respectively. We compare our proposed framework against other domain-adaptive and non-domain-adaptive multi-object tracking baselines using this dataset and show that incorporating domain adaptation at multiple levels of the tracking pipeline yields significant improvements. The code and the dataset are available at https://github.com/chamathabeysinghe/da-tracker.

Categories of Point of Interest (POI) facilitate location-based services from many aspects like location search and POI recommendation. However, POI categories are often incomplete and new POIs are being consistently generated, this rises the demand for semantic annotation for POIs, i.e., labeling the POI with a semantic category. Previous methods usually model sequential check-in information of users to learn POI features for annotation. However, users' check-ins are hardly obtained in reality, especially for those newly created POIs. In this context, we present a Spatial-Textual POI Annotation (STPA) model for static POIs, which derives POI categories using only the geographic locations and names of POIs. Specifically, we design a GCN-based spatial encoder to model spatial correlations among POIs to generate POI spatial embeddings, and an attention-based text encoder to model the semantic contexts of POIs to generate POI textual embeddings. We finally fuse the two embeddings and preserve multi-view correlations for semantic annotation. We conduct comprehensive experiments to validate the effectiveness of STPA with POI data from AMap. Experimental results demonstrate that STPA substantially outperforms several competitive baselines, which proves that STPA is a promising approach for annotating static POIs in map services.

While it has long been recognized that a team of individual learning agents can be greater than the sum of its parts, recent work has shown that larger teams are not necessarily more effective than smaller ones. In this paper, we study why and under which conditions certain team structures promote effective learning for a population of individual learning agents. We show that, depending on the environment, some team structures help agents learn to specialize into specific roles, resulting in more favorable global results. However, large teams create credit assignment challenges that reduce coordination, leading to large teams performing poorly compared to smaller ones. We support our conclusions with both theoretical analysis and empirical results.

Human intelligence comes from the capability to describe and make sense of the world surrounding us, often in a lifelong manner. Online Learning (OL) allows a model to simulate this capability, which involves processing data in sequence, making predictions, and learning from predictive errors. However, traditional OL assumes a fixed set of features to describe data, which can be restrictive. In reality, new features may emerge and old features may vanish or become obsolete, leading to an open feature space. This dynamism can be caused by more advanced or outdated technology for sensing the world, or it can be a natural process of evolution. This paper reviews recent breakthroughs that strived to enable OL in open feature spaces, referred to as Utilitarian Online Learning (UOL). We taxonomize existing UOL models into three categories, analyze their pros and cons, and discuss their application scenarios. We also benchmark the performance of representative UOL models, highlighting open problems, challenges, and potential future directions of this emerging topic.

Artificial Intelligence (AI) has gradually attracted attention in the field of artistic creation, resulting in a debate on the evaluation of AI artistic outputs. However, there is a lack of common criteria for objective artistic evaluation both of human and AI creations. This is a frequent issue in the field of dance, where different performance metrics focus either on evaluating human or computational skills separately. This work proposes a methodological approach for the artistic evaluation of both AI and human artistic creations in the field of robotic dance. First, we define a series of common initial constraints to create robotic dance choreographies in a balanced initial setting, in collaboration with a group of human dancers and choreographer. Then, we compare both creation processes through a human audience evaluation. Finally, we investigate which choreography aspects (e.g., the music genre) have the largest impact on the evaluation, and we provide useful guidelines and future research directions for the analysis of interconnections between AI and human dance creation.

In recent studies, the generalization properties for distributed learning and random features assumed the existence of the target concept over the hypothesis space. However, this strict condition is not applicable to the more common non-attainable case. In this paper, using refined proof techniques, we first extend the optimal rates for distributed learning with random features to the non-attainable case. Then, we reduce the number of required random features via data-dependent generating strategy, and improve the allowed number of partitions with additional unlabeled data. Theoretical analysis shows these techniques remarkably reduce computational cost while preserving the optimal generalization accuracy under standard assumptions. Finally, we conduct several experiments on both simulated and real-world datasets, and the empirical results validate our theoretical findings.

Language models are known to be vulnerable to textual adversarial attacks, which add human-imperceptible perturbations to the input to mislead DNNs. It is thus imperative to devise effective attack algorithms to identify the deficiencies of DNNs before real-world deployment. However, existing word-level attacks have two major deficiencies: (1) They may change the semantics of the original sentence. (2) The generated adversarial sample can appear unnatural to humans due to the introduction of out-of-domain substitute words. In this paper, to address such drawbacks, we propose a semantics- and domain-aware word-level attack method. Specifically, we greedily replace the important words in a sentence with the ones suggested by a language model. The language model is trained to be semantics- and domain-aware via contrastive learning and in-domain pre-training. Furthermore, to balance the quality of adversarial examples and the attack success rate, we propose an iterative updating framework to optimize the contrastive learning loss and the in-domain pre-training loss in circular order. Comprehensive experimental comparisons confirm the superiority of our approach. Notably, compared with state-of-the-art benchmarks, our strategy can achieve over 3\% improvement in attack success rates and 9.8\% improvement in the quality of adversarial examples.

Scene text image super-resolution (STISR), aiming to improve image quality while boosting downstream scene text recognition accuracy, has recently achieved great success. However, most existing methods treat the foreground (character regions) and background (non-character regions) equally in the forward process, and neglect the disturbance from the complex background, thus limiting the performance. To address these issues, in this paper, we propose a novel method LEMMA that explicitly models character regions to produce high-level text-specific guidance for super-resolution. To model the location of characters effectively, we propose the location enhancement module to extract character region features based on the attention map sequence. Besides, we propose the multi-modal alignment module to perform bidirectional visual-semantic alignment to generate high-quality prior guidance, which is then incorporated into the super-resolution branch in an adaptive manner using the proposed adaptive fusion module. Experiments on TextZoom and four scene text recognition benchmarks demonstrate the superiority of our method over other state-of-the-art methods. Code is available at https://github.com/csguoh/LEMMA.

GAN-generated image detection now becomes the first line of defense against the malicious uses of machine-synthesized image manipulations such as deepfakes. Although some existing detectors work well in detecting clean, known GAN samples, their success is largely attributable to overfitting unstable features such as frequency artifacts, which will cause failures when facing unknown GANs or perturbation attacks. To overcome the issue, we propose a robust detection framework based on a novel multi-view image completion representation. The framework first learns various view-to-image tasks to model the diverse distributions of genuine images. Frequency-irrelevant features can be represented from the distributional discrepancies characterized by the completion models, which are stable, generalized, and robust for detecting unknown fake patterns. Then, a multi-view classification is devised with elaborated intra- and inter-view learning strategies to enhance view-specific feature representation and cross-view feature aggregation, respectively. We evaluated the generalization ability of our framework across six popular GANs at different resolutions and its robustness against a broad range of perturbation attacks. The results confirm our method's improved effectiveness, generalization, and robustness over various baselines.

Pruning has been extensively studied in Transformer-based language models to improve efficiency. Typically, we zero (prune) unimportant model weights and train a derived compact model to improve final accuracy. For pruned weights, we treat them as useless and discard them. This usually leads to significant model accuracy degradation. In this paper, we focus on attention head pruning as head attention is a key component of the transformer-based language models and provides interpretable knowledge meaning. We reveal the relationship between pruned attention heads and retained heads and provide a solution to recycle the discarded knowledge from the pruned heads, named peer distillation. We also develop an automatic framework to locate the to-be-pruned attention heads in each layer, freeing the time-consuming human labor in tuning hyperparameters.Experimental results on the General Language Understanding Evaluation (GLUE) benchmark are provided using BERT model. By recycling discarded knowledge from pruned heads, the proposed method maintains model performance across all nine tasks while reducing heads by over 58% on average and outperforms state-of-the-art techniques (e.g., Random, HISP, L0 Norm, SMP).

Recently, Decision Transformer (DT) pioneered the offline RL into a contextual conditional sequence modeling paradigm, which leverages self-attended autoregression to learn from global target rewards, states, and actions. However, many applications have a severe delay of the above signals, such as the agent can only obtain a reward signal at the end of each trajectory. This delay causes an unwanted bias cumulating in autoregressive learning global signals. In this paper, we focused its virtual example on episodic reinforcement learning with trajectory feedback. We propose a new reward redistribution algorithm for learning parameterized reward functions, and it decomposes the long-delayed reward onto each timestep. To improve the redistributing's adaptation ability, we formulate the previous decomposition as a bi-level optimization problem for global optimal. We extensively evaluate the proposed method on various benchmarks and demonstrate an overwhelming performance improvement under long-delayed settings.

Named entity recognition (NER) aims to locate and classify named entities in natural language texts. Most existing high-performance NER models employ a supervised paradigm, which requires a large quantity of high-quality annotated data during training. In order to help NER models perform well in few-shot scenarios, data augmentation approaches attempt to build extra data by means of random editing or by using end-to-end generation with PLMs. However, these methods focus on only the fluency of generated sentences, ignoring the syntactic correlation between the new and raw sentences. Such uncorrelation also brings low diversity and inconsistent labeling of synthetic samples. To fill this gap, we present SAINT (Syntactic-Aware InsertioN Transformer), a hard-constraint controlled text generation model that incorporates syntactic information. The proposed method operates by inserting new tokens between existing entities in a parallel manner. During insertion procedure, new tokens will be added taking both semantic and syntactic factors into account. Hence the resulting sentence can retain the syntactic correctness with respect to the raw data. Experimental results on two benchmark datasets, i.e., Ontonotes and Wikiann, demonstrate the comparable performance of SAINT over the state-of-the-art baselines.

Conversational recommendation systems (CRS) aim to timely and proactively acquire user dynamic preferred attributes through conversations for item recommendation. In each turn of CRS, there naturally have two decision-making processes with different roles that influence each other: 1) director, which is to select the follow-up option (i.e., ask or recommend) that is more effective for reducing the action space and acquiring user preferences; and 2) actor, which is to accordingly choose primitive actions (i.e., asked attribute or recommended item) to estimate the effectiveness of the director’s option. However, existing methods heavily rely on a unified decision-making module or heuristic rules, while neglecting to distinguish the roles of different decision procedures, as well as the mutual influences between them. To address this, we propose a novel Director-Actor Hierarchical Conversational Recommender (DAHCR), where the director selects the most effective option, followed by the actor accordingly choosing primitive actions that satisfy user preferences. Specifically, we develop a dynamic hypergraph to model user preferences and introduce an intrinsic motivation to train from weak supervision over the director. Finally, to alleviate the bad effect of model bias on the mutual influence between the director and actor, we model the director’s option by sampling from a categorical distribution. Extensive experiments demonstrate that DAHCR outperforms state-of-the-art methods.

Optimized trade execution is to sell (or buy) a given amount of assets in a given time with the lowest possible trading cost. Recently, reinforcement learning (RL) has been applied to optimized trade execution to learn smarter policies from market data. However, we find that many existing RL methods exhibit considerable overfitting which prevents them from real deployment. In this paper, we provide an extensive study on the overfitting problem in optimized trade execution. First, we model the optimized trade execution as offline RL with dynamic context (ORDC), where the context represents market variables that cannot be influenced by the trading policy and are collected in an offline manner. Under this framework, we derive the generalization bound and find that the overfitting issue is caused by large context space and limited context samples in the offline setting. Accordingly, we propose to learn compact representations for context to address the overfitting problem, either by leveraging prior knowledge or in an end-to-end manner. To evaluate our algorithms, we also implement a carefully designed simulator based on historical limit order book (LOB) data to provide a high-fidelity benchmark for different algorithms. Our experiments on the high-fidelity simulator demonstrate that our algorithms can effectively alleviate overfitting and achieve better performance.

Mental health is becoming an increasingly prominent health challenge. Despite a plethora of studies analysing and mitigating bias for a variety of tasks such as face recognition and credit scoring, research on machine learning (ML) fairness for mental health has been sparse to date. In this work, we focus on gender bias in mental health and make the following contributions. First, we examine whether bias exists in existing mental health datasets and algorithms. Our experiments were conducted using Depresjon, Psykose and D-Vlog. We identify that both data and algorithmic bias exist. Second, we analyse strategies that can be deployed at the pre-processing, in-processing and post-processing stages to mitigate for bias and evaluate their effectiveness. Third, we investigate factors that impact the efficacy of existing bias mitigation strategies and outline recommendations to achieve greater gender fairness for mental health. Upon obtaining counter-intuitive results on D-Vlog dataset, we undertake further experiments and analyses, and provide practical suggestions to avoid hampering bias mitigation efforts in ML for mental health.

This paper outlines some of the key methods we developed towards the formal verification of multi- agent systems, covering both symbolic and connectionist systems. It discusses logic-based methods for the verification of unbounded multi-agent systems (i.e., systems composed of an arbitrary number of homogeneous agents, e.g., robot swarms), optimisation approaches for establishing the robustness of neural network models, and methods for analysing properties of neuro-symbolic multi-agent systems.

Collaborative tasks often begin with partial task knowledge and incomplete plans from each partner. To complete these tasks, partners need to engage in situated communication with their partners and coordinate their partial plans towards a complete plan to achieve a joint task goal. While such collaboration seems effortless in a human-human team, it is highly challenging for human-AI collaboration. To address this limitation, this paper takes a step towards Collaborative Plan Acquisition, where humans and agents strive to learn and communicate with each other to acquire a complete plan for joint tasks. Specifically, we formulate a novel problem for agents to predict the missing task knowledge for themselves and for their partners based on rich perceptual and dialogue history. We extend a situated dialogue benchmark for symmetric collaborative tasks in a 3D blocks world and investigate computational strategies for plan acquisition. Our empirical results suggest that predicting the partner's missing knowledge is a more viable approach than predicting one's own. We show that explicit modeling of the partner's dialogue moves and mental states produces improved and more stable results than without. These results provide insight for future AI agents that can predict what knowledge their partner is missing and, therefore, can proactively communicate such information to help the partner acquire such missing knowledge toward a common understanding of joint tasks.