Singapore

We investigate the capacity of Large Language Models (LLMs) for imaginative reasoning—the proactive construction, testing, and revision of hypotheses in information-sparse environments. Existing benchmarks, often static or focused on social deduction, fail to capture the dynamic, exploratory nature of this reasoning process. To address this gap, we introduce a comprehensive research framework based on the classic &quot;Turtle Soup&quot; game, integrating a benchmark, an agent, and an evaluation protocol.
We present TurtleSoup-Bench, the first large-scale, bilingual, interactive benchmark for imaginative reasoning, comprising 800 turtle soup stories sourced from both the Internet and expert authors. We also propose Mosaic-Agent, a novel agent designed to assess LLMs&#39; performance in this setting. To evaluate reasoning quality, we develop a multi-dimensional protocol measuring logical consistency, detail completion, and conclusion alignment.
Experiments with leading LLMs reveal clear capability limits, common failure patterns, and a significant performance gap compared to humans. Our work offers new insights into LLMs&#39; imaginative reasoning and establishes a foundation for future research on exploratory agent behavior.

AAAI 2026

What to Ask Next? Probing the Imaginative Reasoning of LLMs with TurtleSoup Puzzles

We investigate the capacity of Large Language Models (LLMs) for imaginative reasoning—the proactive construction, testing, and revision of hypotheses in information-sparse environments. Existing benchmarks, often static or focused on social deduction, fail to capture the dynamic, exploratory nature of this reasoning process. To address this gap, we introduce a comprehensive research framework based on the classic "Turtle Soup" game, integrating a benchmark, an agent, and an evaluation protocol.
We present TurtleSoup-Bench, the first large-scale, bilingual, interactive benchmark for imaginative reasoning, comprising 800 turtle soup stories sourced from both the Internet and expert authors. We also propose Mosaic-Agent, a novel agent designed to assess LLMs' performance in this setting. To evaluate reasoning quality, we develop a multi-dimensional protocol measuring logical consistency, detail completion, and conclusion alignment.
Experiments with leading LLMs reveal clear capability limits, common failure patterns, and a significant performance gap compared to humans. Our work offers new insights into LLMs' imaginative reasoning and establishes a foundation for future research on exploratory agent behavior.

poster

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-26 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.<br><br>

To access this event page, you need to log in with the **email address you registered with**. <br>Access credentials will be sent to your email from Underline -  subject line "Welcome to AAAI 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://aaai.org/conference/aaai/aaai-26/registration/) first.

**Online Registration Form**: https://aaai.getregistered.net/conference-2026 

Registration Required

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

Graph-Level Anomaly Detection (GLAD) seeks to identify anomalous graphs within graph datasets, which has significant applications across diverse real-world fields. Most existing GLAD methods are trained in an unsupervised manner due to high costs for labeling, resulting in sub-optimal performance when compared to supervised methods. To fill this gap, we propose a $\textbf{D}$isentangled $\textbf{G}$eneration-Based $\textbf{P}$rototypical $\textbf{A}$lignment $\textbf{(DGPA)}$ method that extends graph-level anomaly detection to Few-Shot Unsupervised Domain Adaptation (FUDA) setting, aiming to identify anomalous graphs from a set of unlabeled graphs (target domain) by using partially labeled graphs from a different but related domain (source domain), which fulfills the practical requirement of transferring anomaly knowledge. This is specifically achieved through a dedicated Disentangled Sample Generation module, which addresses $\textbf{label scarcity}$ by generating faithful samples with disentangled representation learning grounded in Information Bottleneck principle, along with a Graph-based Prototypical Self-Supervision module, which alleviates $\textbf{domain shift}$ by encoding and aligning semantic structures in the shared latent space across domains in a self-supervised manner. Extensive experiments on five benchmark datasets reveal the effectiveness of our proposed DGPA.

Disentangled Generation-Based Prototypical Alignment for Few-Shot Unsupervised Domain Adaptation in Graph-Level Anomaly Detection

Scene Text Editing (STE) aims to naturally modify text in images while preserving visual consistency, the decisive factors of which can be divided into three parts, i.e., text style, text content, and background. Previous methods have struggled with incomplete disentanglement of editable attributes, typically addressing only one aspect—such as editing text content—thus limiting controllability and visual consistency. To overcome these limitations, we propose TripleFDS, a novel framework for STE with disentangled modular attributes, and an accompanying dataset called SCB Synthesis. 
SCB Synthesis provides robust training data for triple feature disentanglement by utilizing the "SCB Group", a novel construct that combines three attributes per image to generate diverse, disentangled training groups.
Leveraging this construct as a basic training unit, TripleFDS first disentangles triple features, ensuring semantic accuracy through inter-group contrastive regularization and preventing redundancy through intra-sample multi-feature orthogonality. In the synthesis phase, TripleFDS performs feature remapping to prevent "shortcut" phenomena during reconstruction and mitigate potential feature leakage.
Trained on 125,000 SCB Groups, TripleFDS achieves state-of-the-art image fidelity (SSIM of 44.54) and text accuracy (ACC of 93.58\%) on the mainstream STE benchmarks. Besides superior performance, the more flexible editing of TripleFDS supports new operations such as style replacement and background transfer.

TripleFDS: Triple Feature Disentanglement and Synthesis for Scene Text Editing

Federated learning (FL) enables collaborative model training across distributed nodes without exposing raw data, but its decentralized nature makes it vulnerable in trust-deficient environments. Inference attacks may recover sensitive information from gradient updates, while poisoning attacks can degrade model performance or induce malicious behaviors. Existing defenses often suffer from high communication and computation costs, or limited detection precision. To address these issues, we propose LSHFed, a robust and communication-efficient FL framework that simultaneously enhances aggregation robustness and privacy preservation. At its core, LSHFed incorporates LSHGM, a novel gradient verification mechanism that projects high-dimensional gradients into compact binary representations via multi-hyperplane locality-sensitive hashing. This enables accurate detection and filtering of malicious gradients using only their irreversible hash forms, thus mitigating privacy leakage risks and substantially reducing transmission overhead. Extensive experiments demonstrate that LSHFed maintains high model performance even when up to 50% of participants are collusive adversaries, while achieving up to a 1000× reduction in gradient verification communication compared to full-gradient methods.

LSHFed: Robust and Communication-Efficient Federated Learning with Locally-Sensitive Hashing Gradient Mapping

Recent multimodal fusion methods, integrating images with LiDAR point clouds, have shown promise in scene flow estimation. However, the fusion of 4D millimeter wave radar and LiDAR remains unexplored. Unlike LiDAR, radar is cheaper, more robust in various weather conditions and can detect point-wise velocity, making it a valuable complement to LiDAR. However, radar inputs pose challenges due to noise, low resolution, and sparsity. Moreover, there is currently no dataset that combines LiDAR and radar data specifically for scene flow estimation. To address this gap, we construct a Radar-LiDAR scene flow dataset based on a public real-world automotive dataset. We propose an effective preprocessing strategy for radar denoising and scene flow label generation, deriving more reliable flow ground truth for radar points out of the object boundaries. Additionally, we introduce RaLiFlow, the first joint scene flow learning framework for 4D radar and LiDAR, which achieves effective radar-LiDAR fusion through a novel Dynamic-aware Bidirectional Cross-modal Fusion (DBCF) module and a carefully designed set of loss functions. The DBCF module integrates dynamic cues from radar into the local cross-attention mechanism, enabling the propagation of contextual information across modalities. Meanwhile, the proposed loss functions mitigate the adverse effects of unreliable radar data during training and enhance the instance-level consistency in scene flow predictions from both modalities, particularly for dynamic foreground areas. Extensive experiments on the repurposed scene flow dataset demonstrate that our method outperforms existing LiDAR-based and radar-based single-modal methods by a significant margin. Our source code will be made available upon acceptance.

RaLiFlow: Scene Flow Estimation with 4D Radar and LiDAR Point Clouds

Pseudo-random number generators (PRNGs) are high-nonlinear processes, and they are key blocks in optimization of Large language models. Transformers excel at processing complex nonlinear relationships. Thus it is reasonable to generate high-quality pseudo-random numbers based on transformers. In this paper, we explore this question from both theoretical and practical perspectives, highlighting the potential benefits and implications of Transformer in PRNGs. We theoretically demonstrate that decoder-only Transformer models with Chain-of-Thought can simulate both the Linear Congruential Generator (LCG) and Mersenne Twister (MT) PRNGs. Based on this, we conclude that the log-precision decoder-only Transformer can represent non-uniform $\text{AC}^0$. Our simulative theoretical findings are validated through experiments. The random numbers generated by Transformer-based PRNGs successfully pass the majority of NIST tests, whose heat maps exhibit clear statistical randomness. Finally, we assess their capability in prediction attacks.

Transformers in Pseudo-Random Number Generation: A Dual Perspective on Theory and Practice

Speculative decoding accelerates large language model (LLM) inference by using a lightweight drafter to propose multiple tokens, which are then verified in parallel by the base model. While effective in English, existing methods often struggle in multilingual scenarios due to static vocabularies and the lack of language-specific instruction data. To address these limitations, we present AdaSpec, a multilingual speculative decoding framework that dynamically adapts both the drafter and vocabulary at decoding time. AdaSpec generates language-specific instruction data using the LLM itself, enabling training of drafters for low-resource languages. It also constructs adaptive vocabularies tailored to each language's characteristics. In addition, we introduce Multi-SpecBench, a comprehensive multilingual benchmark covering seven languages and seven generation tasks, to evaluate multilingual speculative decoding performance. Extensive experiments show that AdaSpec achieves up to 2.3× speedup over the state-of-the-art method of EAGLE-2, even in English, demonstrating its effectiveness across diverse languages and tasks. The implementation of AdaSpec will be publicly available.

AdaSpec: Adaptive Multilingual Speculative Decoding with Self-Synthesized Language-Aware Training and Vocabulary Simplification

Teachers' emotional states are critical in educational scenarios, profoundly impacting teaching efficacy, student engagement, and learning achievements. However, existing studies often fail to accurately capture teachers' emotions due to the performative nature and overlook the critical impact of instructional information on emotional expression.In this paper, we systematically investigate teacher sentiment analysis by building both the dataset and the model accordingly. We construct the first large-scale teacher multimodal sentiment analysis dataset, T-MED.To ensure labeling accuracy and efficiency, we employ a human-machine collaborative labeling process.The T-MED dataset includes 14,938 instances of teacher emotional data from 250 real classrooms across 11 subjects ranging from K-12 to higher education, integrating multimodal text, audio, video, and instructional information.Furthermore, we propose a novel asymmetric attention-based multimodal teacher sentiment analysis model, AAM-TSA.AAM-TSA introduces an asymmetric attention mechanism and hierarchical gating unit to enable differentiated cross-modal feature fusion and precise emotional classification. Experimental results demonstrate that AAM-TSA significantly outperforms existing state-of-the-art methods in terms of accuracy and interpretability on the T-MED dataset.

Advancing Multimodal Teacher Sentiment Analysis: The Large-Scale T-MED Dataset & the Effective AAM-TSA Model

Significant efforts have been focused on enhancing the utilization of multiple node features and topological structures in multi-view graph learning through explicit model-based and implicit deep-based methodologies. The former excel in embedding prior knowledge in learning graphs, thereby offering the theory-level interpretability but limiting in the application-level flexibility because of manual parameter selection. In contrast, the latter leverage automatic differentiation mechanisms for learning graphs, providing greater application-level flexibility but suffering from reduced the theory-level interpretability due to their opaque nature. Motivated by these observations, we propose an interpretable deep unfolding network for mutual-benefit multi-view graph learning, aiming to combine the strengths of both approaches. First, we employ the ADMM optimizer to effectively solve the multi-view graph learning model with sparse and low-rank constraints, integrating this iterative solution of mathematical support into the construction of explicit-level deep unfolding networks, thereby enhancing the theory-level interpretability. Second, we convert certain conditions into implicit-level losses and leverage automatic differentiation capabilities to update parameters, thereby reducing the necessity for manual parameter tuning and enhancing the application-level flexibility. Through this collaboration, we optimize multi-view learning for a graph representation that balances interpretability and flexibility. Empirical evaluations across six diverse datasets demonstrate the effectiveness and superiority of the proposed method compared to state-of-the-art approaches.

Graph Meets Deep Unfolding: An Interpretable Mutual-benefit Multi-view Learning Network

The abundance of fine-grained spatio-temporal data, such as traffic sensor networks, offers vast opportunities for scientific discovery. However, inferring causal relationships from such observational data remains challenging, particularly due to unobserved confounders that are specific to units (e.g., geographical locations) yet influence outcomes over time. Most existing methods for spatio-temporal causal inference assume that all confounders are observed, an assumption that is often violated in practice. In this paper, we introduce Spatio-Temporal Hierarchical Causal Models (ST-HCMs), a novel graphical framework that extends hierarchical causal modeling to the spatio-temporal domain. At the core of our approach is the Spatio-Temporal Collapse Theorem, which shows that a complex ST-HCM converges to a simpler flat causal model as the amount of subunit data increases. This theoretical result enables a general procedure for causal identification, allowing ST-HCMs to recover causal effects even in the presence of unobserved, time-invariant unit-level confounders, a scenario where standard non-hierarchical models fail. We validate the effectiveness of our framework on both synthetic and real-world datasets, demonstrating its potential for robust causal inference in complex dynamic systems.

Spatio-Temporal Hierarchical Causal Models

Graph OOD detection is crucial in open-world scenarios, where OOD samples may manifest in diverse forms such as open-set deviations, feature-similar shifts, and structural anomalies, each exhibiting distinct geometric characteristics. However, most existing methods adopt a one-size-fits-all geometric assumption (typically Euclidean space), which inadequately captures the diverse nature of real-world distribution shifts. Therefore, adaptively selecting geometric spaces according to the properties of OOD samples is critical for their effective representation and reliable identification. Motivated by this, we revisit the graph OOD detection task under diverse distribution shifts and propose UniGOD, a unified framework serving as a graph foundation model for this task. UniGOD comprises two core modules: GeoUP and DynEVO. GeoUP module adaptively perceives the geometric space (such as Euclidean, hyperbolic, and hyperspherical space) by learning the curvature $\kappa$ of Riemannian manifolds. DynEVO module leverages the dynamic nature of neural SDEs to reveal pronounced uncertainty differences between ID/OOD samples, which are reflected in the divergent evolutionary trajectories of node embeddings induced by $\kappa$-GNN iterations. With the geometry-dynamics coupling mechanism of the above two modules, UniGOD effectively captures the diverse distribution shifts. Extensive experiments demonstrate its superior performance over existing SOTA methods.

Premium content

Next from AAAI 2026

Disentangled Generation-Based Prototypical Alignment for Few-Shot Unsupervised Domain Adaptation in Graph-Level Anomaly Detection

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES