IJCAI 2024 Papers

This paper demonstrates a novel method for LiDAR-based 3D object detection, addressing major field challenges: sparsity and occlusion. Our approach leverages temporal point cloud sequences to generate frames that provide comprehensive views of objects from multiple angles. To address the challenge of generating these frames in real-time, we employ Knowledge Distillation within a Teacher-Student framework, allowing the Student model to emulate the Teacher’s advanced perception. We pioneered the application of weak-to-strong generalization in computer vision by training our Teacher model on enriched, object-complete data. In this demo, we showcase the exceptional quality of labels produced by the X-Ray Teacher on object-complete frames, showing our method distilling its knowledge to enhance object 3D detection models.

The challenge in constructing artificial social agents is to enable adaptation ability to novel agents, and is called zero-shot coordination (ZSC). A promising approach is to train the adaptive agents by interacting with a diverse pool of collaborators, assuming that the greater the diversity in other agents seen during training, the better the generalisation. In this paper, we explore an alternative procedure by considering the behavioural predictability of collaborators, i.e. whether their actions and intentions are predictable, and use it to select a diverse set of agents for the training pool. More specifically, we develop a pool of agents through self-play training during which agents' behaviour evolves and has diversity in levels of behavioural predictability (LoBP) through its evolution. We construct an observer to compute the level of behavioural predictability for each version of the collaborators. To do so, the observer is equipped with the theory of mind (ToM) capability to learn to infer the actions and intentions of others. We then use an episodic memory based on the LoBP metric to maintain agents with different levels of behavioural predictability in the pool of agents. Since behaviours that emerge at the later training phase are more complex and meaningful, the memory is updated with the latest versions of training agents. Our extensive experiments demonstrate that LoBP-based diversity training leads to better ZSC than other diversity training methods.

Offline Reinforcement Learning (RL) aims to learn policies from pre-collected datasets that capture only a subset of the environment's dynamics. The predominant approach has been to solve a constrained optimization formulation, which ensures that the policy visits state-action pairs within the support of the offline dataset. However, this approach has limited the ability to make decisions when the agent faces unknown parts of the environment at deployment time. To address the challenge of decision-making in out-of-support regions, model-based Bayes-adaptive approaches have been proposed by considering all dynamics models that could potentially be the true environment. Since it is generally infeasible to compute the posterior of all dynamics models based on the offline dataset, these approaches usually approximate the posterior by using a finite ensemble of highly probable dynamics models. Hence, the diversity of these models is the key to obtaining good policies. In this work, we propose MoDAP (Model-based Diverse Adaptive Policy Learning), an algorithm to enable the adaptive policy to make informed decisions in previously unexplored states. MoDAP adopts an iterative strategy that simultaneously training the policy and dynamics models. The policy optimization seeks to maximize expected returns across dynamics models, while the dynamics models are trained to promote policy diversification through the proposed information-theoretic objective. We evaluate MoDAP through experiments on the D4RL and NeoRL benchmarks, showcasing its performance superiority over state-of-the-art algorithms.

Automatic cell type annotation aims to transfer the label knowledge from label-abundant reference data to label-scarce target data, which makes encouraging progress in single-cell RNA-seq data analysis. While previous works have focused on classifying close-set cells and detecting open-set cells during testing, it is still essential to be able to classify unknown cell types as human beings. Additionally, few efforts have been devoted to addressing the challenge of common long-tail dilemma in cell type annotation data. Therefore, in this paper, we propose an innovative distribution-independent universal cell type identification framework called scDET from the perspective of autonomously equilibrated dual-consultative contrastive learning. Our model can generate fine-grained predictions for both close-set and open-set cell types in a long-tailed open-world environment. scDET consists of a contrastive-learning branch and a pseudo-labeling branch, which work collaboratively to provide interactive supervision. Specifically, the contrastive-learning branch provides reliable distribution estimation to regularize the predictions of the pseudo-labeling branch, which in turn guides itself through self-balanced knowledge transfer and a designed novel soft contrastive loss. Extensive experimental results on various evaluation datasets demonstrate the superior performance of scDET over other state-of-the-art single-cell clustering and annotation methods.

The fabrication of visual misinformation on the web and social media has increased exponentially with the advent of foundational text-to-image diffusion models. Namely, Stable Diffusion inpainters allow the synthesis of maliciously inpainted images of personal and private figures, and copyrighted contents, also known as deepfakes. To combat such generations, a disruption framework, namely Photoguard, has been proposed, where it adds adversarial noise to the context image to disrupt their inpainting synthesis. While their framework suggested a diffusion-friendly approach, the disruption is not sufficiently strong and it requires a significant amount of GPU and time to immunize the context image. In our work, we re-examine both the minimal and favorable conditions for a successful inpainting disruption, proposing DDD, a “Digression guided Diffusion Disruption” framework. First, we identify the most adversarially vulnerable diffusion timestep range with respect to the hidden space. Within this scope of noised manifold, we pose the problem as a semantic digression optimization. We maximize the distance between the inpainting instance's hidden states and a semantic-aware hidden state centroid, calibrated both by Monte Carlo sampling of hidden states and a discretely projected optimization in the token space. Effectively, our approach achieves stronger disruption and a higher success rate than Photoguard while lowering the GPU memory requirement, and speeding the optimization up to three times faster.

Modeling time series data has become a very at tractive research topic due to its wide application, such as human activity recognition, financial forecasting and sensor-based automatic system monitoring. Recently deep learning models have shown great advances in modeling the time series data but they heavily depend on a large amount of labeled data. To avoid costly labeling, this paper explores domain adaptation from a labeled source domain to the unlabeled target domain on time series data. To achieve the goal, we propose a disentangled representation learning framework named CADT to disentangle the domain-invariant features from the domain-specific ones. Particularly, CADT is injected with a novel class-wise hypersphere loss to improve the generalization of the classifier from the source domain to the target domain. Intuitively, it restricts the source data of the same class within the same hypersphere and minimizes the radius of it, which in turn enlarges the margin between different classes and makes the decision boundary of both domains easier. We further devise several kinds of domain-preserving data augmentation methods to better capture the domain-specific patterns. Extensive experiments on two public datasets and two real-world applications demonstrate the effectiveness of the proposed model against several state-of-the-art baselines.

The emergence of AIGC has brought attention to the issue of generating realistic deceptive content. While AIGC has the potential to revolutionize content creation, it also facilitates criminal activities. Specifically, the manipulation of speech has been exploited in tele-fraud and financial fraud schemes, posing a significant threat to societal security. Current deep learning-based methods for detecting forged speech extract mixed features from the original speech, which often contain redundant information. Moreover, these methods fail to consider the distinct characteristics of human voice-specific features and the diversity of background environmental sounds. This paper introduces a framework called Discriminative fEature dEcoupling enhanceMent (DEEM) for detecting speech forgery. Initially, the framework decouples the original speech into human voice features and background sound features. Subsequently, DEEM enhances voice-specific features through temporal dimension aggregation and improves continuity-related features in the background sound map via spectral-dimension aggregation. By employing the decoupling enhancement features, extensive experiments demonstrate that DEEM achieves an accuracy improvement of over 5% on FoR dataset compared to the state-of-the-art methods.

In modern federated learning, one of the main challenges is to account for inherent heterogeneity and the diverse nature of data distributions for different clients. This problem is often addressed by introducing personalization of the models towards the data distribution of the particular client. However, a personalized model might be unreliable when applied to the data that is not typical for this client. Eventually, it may perform worse for these data than the non-personalized global model trained in a federated way on the data from all the clients. This paper presents a new approach to federated learning that allows selecting a model from global and personalized ones that would perform better for a particular input point. It is achieved through a careful modeling of predictive uncertainties that helps to detect local and global in- and out-of-distribution data and use this information to select the model that is confident in a prediction. The comprehensive experimental evaluation on the popular real-world image datasets shows the superior performance of the model in the presence of out-of-distribution data while performing on par with state-of-the-art personalized federated learning algorithms in the standard scenarios.

We present a novel prompting strategy for artificial intelligence driven digital avatars. To better quantify how our prompting strategy affects anthropomorphic features like humor, authenticity, and favorability we present Crowd Vote - an adaptation of Crowd Score that allows for judges to elect a large language model (LLM) candidate over competitors answering the same or similar prompts. To visualize the responses of our LLM, and the effectiveness of our prompting strategy we propose an end-to-end framework for creating high-fidelity artificial intelligence (AI) driven digital avatars. This pipeline effectively captures an individual's essence for interaction and our streaming algorithm delivers a high-quality digital avatar with real-time audio-video streaming from server to mobile device. Both our visualization tool, and our Crowd Vote metrics demonstrate our AI driven digital avatars have state-of-the-art humor, authenticity, and favorability outperforming all competitors and baselines. In the case of our Donald Trump and Joe Biden avatars, their authenticity and favorability are rated higher than even their real-world equivalents.

Toon shading is a type of non-photorealistic rendering task in animation. Its primary purpose is to render objects with a flat and stylized appearance. As diffusion models have ascended to the forefront of image synthesis, this paper delves into an innovative form of toon shading based on diffusion models, aiming to directly render photorealistic videos into anime styles. In video stylization, existing methods encounter persistent challenges, notably in maintaining consistency and achieving high visual quality. In this paper, we model the toon shading problem as four subproblems, i.e., stylization, consistency enhancement, structure guidance, and colorization. To address the challenges in video stylization, we propose an effective toon shading approach called Diffutoon. Diffutoon is capable of rendering remarkably detailed, high-resolution, and extended-duration videos in anime style. It can also edit the video content according to input prompts via an additional branch. The efficacy of Diffutoon is evaluated through quantitive metrics and human evaluation. Notably, Diffutoon surpasses both open-source and closed-source baseline approaches in our experiments. Our work is accompanied by the release of both the source code and example videos on Github.

Most existing visual-language trackers greatly rely on the initial language descriptions on a target object to extract their multi-modal features. However, the initial language descriptions are often inaccurate in a highly time-varying video sequence and thus greatly deteriorate their tracking performance due to the low quality of extracted multi-modal features. To address this challenge, we propose a Diffusion Mask-Driven Visual-language Tracker (DMTrack) based on a diffusion model. Confronting the issue of low-quality multi-modal features due to inaccurate language descriptions, we leverage the diffusion model to capture high-quality semantic information from multi-modal features and transform it into target mask features. During the training phase, we further enhance the diffusion model's perception of pixel-level features by calculating the loss between the target mask features and the ground truth masks. Additionally, we perform joint localization of the target using both target mask features and visual features, instead of relying solely on multi-modal features for localization. Through extensive experiments on four tracking benchmarks (i.e., LaSOT, TNL2K, LaSOText, and OTB-Lang), we validate that our proposed Diffusion Mask-Driven Visual-language Tracker can improve the robustness and effectiveness of the model.

Traditional image steganography focuses on concealing one image within another, aiming to avoid steganalysis by unauthorized entities. Coverless image steganography (CIS) enhances imperceptibility by not using any cover image. Recent works have utilized text prompts as keys in CIS through diffusion models. However, this approach faces three challenges: invalidated when private prompt is guessed, crafting public prompts for semantic diversity, and the risk of prompt leakage during frequent transmission. To address these issues, we propose DiffStega, an innovative training-free diffusion-based CIS strategy for universal application. DiffStega uses a password-dependent reference image as an image prompt alongside the text, ensuring that only authorized parties can retrieve the hidden information. Furthermore, we develop Noise Flip technique to further secure the steganography against unauthorized decryption. To comprehensively assess our method across general CIS tasks, we create a dataset comprising various image steganography instances. Experiments indicate substantial improvements in our method over existing ones, particularly in aspects of versatility, password sensitivity, and recovery quality. Codes are available at https://github.com/evtricks/DiffStega.

Recent years have witnessed the success of generative adversarial networks and diffusion models in multi-model trajectory prediction. However, prevailing algorithms only explicitly consider human interaction, but ignore the modeling of human intention, yielding that the generated results deviate largely from real trajectories in some complex scenes. In this paper, we analyze the conditions of multi-modal trajectory prediction from two objective perspectives and propose a novel end-to-end framework based on the diffusion model to predict more precise and socially-acceptable trajectories for humans. Firstly, a spatial-temporal aggregation module is built to extract the extrinsic interaction features for capturing socially-acceptable behaviors. Secondly, we explicitly construct the intrinsic intention module to obtain intention features for precise prediction. Finally, we estimate a noise trajectory distribution with these two features as the initiation of diffusion model and leverage denoising process to obtain the final trajectories. Furthermore, to reduce the noise of the initiative trajectory estimation, we present a novel sample consistency loss to constrain multiple predictions. Extensive experiments demonstrate that our method outperforms the state-of-the-art methods on ETH-UCY and SDD benchmarks, specifically achieving 19.0%/24.2% ADE/FDE improvement on ETH-UCY.

Real-time engagement estimation has been an important research topic in human-computer interaction in recent years. The emergence of the NOvice eXpert Interaction (NOXI) dataset, enriched with frame-wise engagement annotations, has catalyzed a surge in research efforts in this domain. Existing feature sequence partitioning methods for ultra-long videos have encountered challenges including insufficient information utilization and repetitive inference. Moreover, those studies focus mainly on the target participants’ features without taking into account those of the interlocutor. To address these issues, we propose the center-based sliding window method to obtain feature subsequences. The core of these subsequences is modeled using our innovative Central Engagement Attention Model (CEAM). Additionally, we introduce the dialogue cross-enhanced module that effectively incorporates the interlocutor’s features via cross-attention. Our proposed method outperforms the current best model, achieving a substantial gain of 1.5% in coordination correlation coefficient (CCC) and establishing a new state-of-the-art result. Our source codes and model checkpoints are available at https://github.com/wujiekd/Dialogue-Cross-Enhanced-CEAM.

An important question in elections is determining whether a candidate can be a winner when some votes are absent. We study this determining winner with absent votes (WAV) problem with elections that take top-truncated ballots. We show that the WAV problem is NP-complete for single transferable vote, Maximin, and Copeland, and propose a special case of positional scoring rule such that the problem can be computed in polynomial time. Our results for top-truncated rankings differ from the results in full rankings as their hardness results still hold when the number of candidates or the number of missing votes are bounded, while we show that the problem can be solved in polynomial time in either case.

Object detection tasks, crucial in safety-critical systems like autonomous driving, focus on pinpointing object locations. These detectors are known to be susceptible to backdoor attacks. However, existing backdoor techniques have primarily been adapted from classification tasks, overlooking deeper vulnerabilities specific to object detection. This paper is dedicated to bridging this gap by introducing Detector Collapse (DC), a brand-new backdoor attack paradigm tailored for object detection. DC is designed to instantly incapacitate detectors (i.e., severely impairing detector's performance and culminating in a denial-of-service). To this end, we develop two innovative attack schemes: Sponge for triggering widespread misidentifications and Blinding for rendering objects invisible. Remarkably, we introduce a novel poisoning strategy exploiting natural objects, enabling DC to act as a practical backdoor in real-world environments. Our experiments on different detectors across several benchmarks show a significant improvement (~10%-60% absolute and ~2-7x relative) in attack efficacy over state-of-the-art attacks.

Large Language Models (LLMs), characterized by being trained on broad amounts of data in a self-supervised manner, have shown impressive performance across a wide range of tasks. Indeed, their generative abilities have aroused interest on the application of LLMs across a wide range of contexts. However, neural networks in general, and LLMs in particular, are known to be vulnerable to adversarial attacks, where an imperceptible change to the input can mislead the output of the model. This is a serious concern that impedes the use of LLMs on high-stakes applications, such as healthcare, where a wrong prediction can imply serious consequences. Even though there are many efforts on making LLMs more robust to adversarial attacks, there are almost no works that study how and where these vulnerabilities that make LLMs prone to adversarial attacks happen. Motivated by these facts, we explore how to localize and understand vulnerabilities, and propose a method, based on Mechanistic Interpretability (MI) techniques, to guide this process. Specifically, this method enables us to detect vulnerabilities related to a concrete task by (i) obtaining the subset of the model that is responsible for that task, (ii) generating adversarial samples for that task, and (iii) using MI techniques together with the previous samples to discover and understand the possible vulnerabilities. We showcase our method on a pretrained GPT-2 Small model carrying out the task of predicting 3-letter acronyms to demonstrate its effectiveness on locating and understanding concrete vulnerabilities of the model.

Detecting abrupt changes in data distribution is one of the most significant tasks in streaming data analysis. Although many unsupervised Change-Point Detection (CPD) methods have been proposed recently to identify those changes, they still suffer from missing subtle changes, poor scalability, or/and sensitivity to outliers. To meet these challenges, we are the first to generalise the CPD problem as a special case of the Change-Interval Detection (CID) problem. Then we propose a CID method, named iCID, based on a recent Isolation Distributional Kernel (IDK). iCID identifies the change interval if there is a high dissimilarity score between two non-homogeneous temporal adjacent intervals. The data-dependent property and finite feature map of IDK enabled iCID to efficiently identify various types of change-points in data streams with the tolerance of outliers. Moreover, the proposed online and offline versions of iCID have the ability to optimise key parameter settings. The effectiveness and efficiency of iCID have been systematically verified on both synthetic and real-world datasets.

This study explores the challenge of sentence-level AI-generated text detection within human-AI collaborative hybrid texts (abbreviated as hybrid texts). Existing studies of AI-generated text detection for hybrid texts often rely on synthetic datasets. These typically involve hybrid texts with a limited number of boundaries, e.g., single-boundary hybrid texts that begin with human-written content and end with machine-generated continuations. We contend that studies of detecting AI-generated content within hybrid texts should cover different types of hybrid texts generated in realistic settings to better inform real-world applications. Therefore, our study utilizes the CoAuthor dataset, which includes diverse, realistic hybrid texts generated through the collaboration between human writers and an intelligent writing system in multi-turn interactions. We adopt a two-step, segmentation-based pipeline: (i) detect segments within a given hybrid text where each segment contains sentences of consistent authorship, and (ii) classify the authorship of each identified segment. Our empirical findings highlight (1) detecting AI-generated sentences in hybrid texts is overall a challenging task because (1.1) human writers' selecting and even editing AI-generated sentences based on personal preferences adds difficulty in identifying the authorship of segments; (1.2) the frequent change of authorship between neighboring sentences within the hybrid text creates difficulties for segment detectors in identifying authorship-consistent segments; (1.3) the short length of text segments within hybrid texts provides limited stylistic cues for reliable authorship determination; (2) before embarking on the detection process, it is beneficial to assess the average length of segments within the hybrid text. This assessment aids in deciding whether (2.1) to employ a text segmentation-based strategy for hybrid texts with longer segments, or (2.2) to adopt a direct sentence-by-sentence classification strategy for those with shorter segments.