AAAI 2026 Papers

Spatiotemporal analysis of facial behavior is a crucial method for evaluating the mental state of depression patients. However, in practice, depressed patients often display facial behaviors similar to healthy individuals due to masking tendencies. Additionally, facial expressions among depressed patients are also different, increasing the difficulty of assessment. To address this, we propose a video-based automatic depression assessment model Dep-MAP for complex facial behaviors of depression patients. Dep-MAP adopts a dual-branch architecture to extract visual features of facial behavior and capture corresponding emotional semantic features. Specifically, the extracted deep semantic features are clustered, resulting in semantically distinct prototype sets, where each severity group learns a set of discriminative facial behavior prototype representations, to suppress inter-class semantic confusion. Subsequently, we propose a semantic prototype-supervised contrastive learning method, which aligns latent semantics between shallow and deep features, realizing emotional semantic guidance and self-knowledge distillation for the visual feature branch, effectively suppressing intra-class difference. Then, we integrate key depression cues across multiple spatiotemporal scales via a multi-scale weighted fusion strategy, achieving automatic depression assessment. Experimental results demonstrate that Dep-MAP effectively identifies potential key frames in temporal sequences, and aggregates key frame representations with semantic consistency, achieving significantly superior state-of-the-art results on the AVEC2013 and AVEC2014 public datasets.

Current brain-computer interfaces primarily decode single motor variables, limiting natural control requiring simultaneous multi-dimensional extraction. We introduce Multi-dimensional Neural Decoding (MND), a task that simultaneously extracts multiple motor variables (direction, position, velocity, acceleration) from single neural population recordings. MND faces two key challenges: cross-task interference when decoding correlated motor dimensions from shared cortical representations, and generalization issues across sessions, subjects, and paradigms. To address these challenges, we propose OrthoSchema, a multi-task framework inspired by cortical orthogonal subspace organization and cognitive schema reuse. OrthoSchema enforces representation orthogonality to eliminate cross-task interference and employs selective feature reuse transfer for few-shot cross-session, subject and paradigm adaptation. Experiments on macaque motor cortex datasets demonstrate that OrthoSchema significantly improves decoding accuracy in cross-session, subject and paradigm generalization tasks, with larger performance improvements when fine-tuning samples are limited. Ablation studies confirm the synergistic effects of all components are crucial, with OrthoSchema effectively modeling cross-task features and capturing session relationships for robust transfer. Our results provide new insights into scalable and robust neural decoding for real-world BCI applications.

Autonomous agents play a crucial role in advancing Artificial General Intelligence, enabling problem decomposition and tool orchestration through Large Language Models (LLMs). However, existing paradigms face a critical trade-off. On one hand, reusable fixed workflows require manual reconfiguration upon environmental changes; on the other hand, flexible reactive loops fail to distill reasoning progress into transferable structures. We introduce Hierarchical Variable Agent (HiVA), a novel framework modeling agentic workflows as self-organized graphs with the Semantic-Topological Evolution (STEV) algorithm, which optimizes hybrid semantic-topological spaces using textual gradients as discrete-domain surrogates for backpropagation. The iterative process comprises Multi-Armed Bandit-infused forward routing, diagnostic gradient generation from environmental feedback, and coordinated updates that co-evolve individual semantics and topology for collective optimization in unknown environments. Experiments on dialogue, coding, Long-context Q&A, mathematical, and agentic benchmarks demonstrate improvements of 5-10% in task accuracy and enhanced resource efficiency over existing baselines, establishing HiVA's effectiveness in autonomous task execution.

Emotional support is a core capability in human-AI interaction, with applications including psychological counseling, role play, and companionship. However, existing evaluations of large language models (LLMs) often rely on short, static dialogues and fail to capture the dynamic and long-term nature of emotional support. To overcome this limitation, we shift from snapshot-based evaluation to trajectory-based assessment, adopting a user-centered perspective that evaluates models based on their ability to improve and stabilize user emotional states over time. Our framework constructs a large-scale benchmark consisting of 328 emotional contexts and 1,152 disturbance events, simulating realistic emotional shifts under evolving dialogue scenarios. To encourage psychologically grounded responses, we constrain model outputs using validated emotion regulation strategies such as situation selection and cognitive reappraisal. User emotional trajectories are modeled as a first-order Markov process, and we apply causally-adjusted emotion estimation to obtain unbiased emotional state tracking. Based on this framework, we introduce three trajectory-level metrics: Baseline Emotional Level (BEL), Emotional Trajectory Volatility (ETV), and Emotional Centroid Position (ECP). These metrics collectively capture user emotional dynamics over time and support comprehensive evaluation of long-term emotional support performance of LLMs. Extensive evaluations across a diverse set of LLMs reveal significant disparities in emotional support capabilities and provide actionable insights for model development.

Generating human motion in complex 3D scenes from text is a challenging task with broad applications. However, existing methods often overlook realistic physical contact, resulting in visually plausible but physically unrealistic motion, e.g., penetration. To alleviate this, we propose IntentMotion, a novel framework that generates human motion in 3D scenes from natural language instructions by explicitly modeling intent. We first introduce the Intention-Guided Contact Field (IGCF). This differentiable voxel-based contact region representation explicitly aligns parsed language roles with spatial contact regions through a hierarchical attention mechanism. IGCF is jointly trained with a diffusion-based motion generator, allowing contact predictions to adapt dynamically through gradient feedback. To improve the controllability and physics-aware motion, we further propose an Intention-Aware Diffusion Model (IADM), which decouples the high-level semantic planning from the low-level contact refinement in a coarse-to-fine process. The optimized contact cues are utilized to guide the synthesis of a coarse trajectory, followed by refining detailed pose sequences under IGCF supervision. Experiments on the HUMANISE and LINGO datasets demonstrate that our IntentMotion outperforms recent baselines in contact accuracy, semantic alignment, and generalization to unseen scenes.

Spiking neural networks (SNNs) have demonstrated significant potential in real-time multi-sensor perception tasks due to their event-driven and parameter-efficient characteristics. A key challenge is the timestep-wise iterative update of neuronal hidden states (membrane potentials), which complicates the trade-off between accuracy and latency. SNNs tend to achieve better performance with longer timesteps, inevitably resulting in higher computational overhead and latency compared to artificial neural networks (ANNs). Moreover, many recent advances in SNNs rely on architecture-specific optimizations, which, while effective with fewer timesteps, often limit generalizability and scalability across modalities and models. To address these limitations, we propose Activation-wise Membrane Potential Propagation (AMP2), a unified hidden state update mechanism for SNNs. Inspired by the spatial propagation of membrane potentials in biological neurons, AMP2 enables dynamic transmission of membrane potentials among spatially adjacent neurons, facilitating spatiotemporal integration and cooperative dynamics of hidden states, thereby improving efficiency and accuracy while reducing reliance on extended temporal updates. This simple yet effective strategy significantly enhances SNN performance across various architectures, including MLPs and CNNs for point cloud and event-based data. Furthermore, ablation studies integrating AMP2 into Transformer-based SNNs for classification tasks demonstrate its potential as a general-purpose and efficient solution for spiking neural networks.

Memory behavior modeling seeks to predict individual recall performance and understand its underlying cognitive mechanisms. However, the dynamic and heterogeneous nature of memory data poses significant challenges to the generalization ability of models under unseen conditions. To address this challenge, we propose an invariant representation learning framework I-Mem that integrates self-supervised contrastive learning with decorrelation constraints, enabling the adaptive identification and suppression of environment-related factors in sequential behavioral data, thereby mitigating the influence of spurious features and enhancing the modeling of stable cognitive structures. Importantly, the method does not rely on explicit environment partitioning or predefined environment labels, while our theoretical analysis demonstrates that it can effectively resist environmental perturbations and facilitate the extraction of invariant structural representations, thereby ensuring adaptability and generalization. Empirical evaluations on both synthetic and real-world datasets further confirm its superiority over mainstream methods in terms of generalization performance and stable feature identification. Feature attribution analysis reveals that I-Mem extracts invariant features aligned with classical cognitive effects, and reflects short-term behavioral patterns that may indicate latent cognitive mechanisms beyond existing theories, highlighting both interpretability and discovery potential.

Collage is a powerful medium for visual expression, traditionally demanding significant artistic expertise and manual effort. Existing methods often struggle with a trade-off between semantic expression and the visual fidelity of the constituent images. To address this, we introduce SCORE (Semantic Collage by Optimizing Rendered Elements), a novel text-driven framework that automates the creation of semantically rich and structurally sound collages. Our key innovation is to shift the optimization process entirely into the image space. By employing a differentiable renderer, we can backpropagate gradients from a powerful, pre-trained text-to-image model directly to the spatial parameters, including position, rotation, and scale, of each image element. We leverage Variational Score Distillation (VSD) to provide robust semantic guidance from a text prompt, ensuring the final layout aligns with the desired concept. Crucially, our ''minimal editing'' principle preserves the integrity of the original elements by forgoing any content-level modifications. The layout is refined by a joint loss function that combines the VSD-based semantic loss with structural regularizers that penalize overlap and enforce boundary constraints. The output of SCORE is a parametric, structured representation that allows further editing and downstream use. Our work reduces the barrier to creative expression and provides a new, powerful paradigm for organizing visual contents.

The recent advancement of Multimodal Large Language Models (MLLMs) is transforming human-computer interaction (HCI) from surface-level exchanges into more nuanced and emotionally intelligent communication. To realize this shift, emotion understanding becomes essential allowing systems to capture subtle cues underlying user intent. Furthermore, providing faithful explanations for predicted emotions is crucial to ensure interpretability and build user trust. However, current MLLM-based methods often generate emotion explanations that diverge from the ground-truth (GT) labels and sometimes even contradict their own predicted emotions. This inconsistency poses a critical risk for misunderstanding and erodes reliability in interactive settings. To address this, we propose a novel approach: the Emotional Rationale Verifier (ERV) and an Explanation Reward. Our method guides the model to produce reasoning that is explicitly consistent with the GT emotion during multimodal emotion recognition without modifying the model architecture or requiring paired video–description annotations. Our method significantly improves faithful explanation–prediction consistency and explanation emotion accuracy on the MAFW and DFEW datasets. Through extensive experiments and human evaluations, we show that our approach not only enhances alignment between explanation and prediction but also empowers MLLMs to deliver emotionally coherent, trustworthy interactions, marking a key step toward truly human-like HCI systems.

We introduce a biologically inspired, multilayer neural architecture composed of Rectified Spectral Units (ReSUs). Each ReSU projects a recent window of its input history onto a canonical direction obtained via canonical correlation analysis (CCA) of previously observed past–future input pairs, and then rectifies either its positive or negative component. By encoding canonical directions in synaptic weights and temporal filters, ReSUs implement a local, self-supervised algorithm for progressively constructing increasingly complex features. To evaluate both computational power and biological fidelity, we trained a two-layer ReSU network in a self-supervised regime on translating natural scenes. First-layer units, each driven by a single pixel, developed temporal filters resembling those of Drosophila post-photoreceptor neurons (L1/L2 and L3), including their empirically observed adaptation to signal-to-noise ratio (SNR). Second-layer units, which pooled spatially over the first layer, became direction-selective—analogous to T4 motion-detecting cells—with learned synaptic weight patterns approximating those derived from connectomic reconstructions. Together, these results suggest that ReSUs offer (i) a principled framework for modeling sensory circuits and (ii) a biologically grounded, backpropagation-free paradigm for constructing deep self-supervised neural networks.

The mapping from sound to neural activity that underlies hearing is highly non-linear. The first few stages of this mapping in the cochlea have been modelled successfully, initially with biophysical models built by hand and, more recently, with DNN models trained on datasets simulated by the biophysical models. Modelling the auditory brain has been a challenge because central auditory processing is too complex for models to be built by hand, and datasets for training DNN models directly have not been available. Recent work has taken advantage of large-scale high resolution neural recordings from the auditory midbrain to build a DNN model of normal hearing with great success. But this model assumes that auditory processing is the same in all brains, and therefore it cannot capture the widely varying effects of hearing loss. We propose a novel variational-conditional model to learn to encode the space of hearing loss directly from recordings of neural activity in the auditory midbrain of normal and noise exposed animals. With hearing loss parametrised by only 6 free parameters per animal, our model accurately predicts 62% of the explainable variance in neural responses from normal hearing animals and 68% for hearing impaired animals, comparable to state of the art animal specific models. We demonstrate that the model can be used to simulate realistic activity from out of sample animals by fitting only the learned conditioning parameters with Bayesian optimisation, achieving crossentropy loss within 2% of the optimum in 15-30 iterations. Including more animals in the training data slightly improved the performance on unseen animals. This model will enable future development of parametrised hearing loss compensation models trained to directly restore normal neural coding in hearing impaired brains, which can be quickly fitted for a new user by human in the loop optimisation.

Music editing is an important step in music production, which has broad applications, including game development and film production. Most existing zero-shot text-guided editing methods rely on pretrained diffusion models by involving forward-backward diffusion processes. However, these methods often struggle to preserve the musical content. Additionally, text instructions alone usually fail to accurately describe the desired music. In this paper, we propose two music editing methods that improve the consistency between the original and edited music by leveraging score distillation. The first method, SteerMusic, is a coarse-grained zero-shot editing approach using delta denoising score. The second method, SteerMusic+, enables fine-grained personalized music editing by manipulating a concept token that represents a user-defined musical style. SteerMusic+ allows for the editing of music into user-defined musical styles that cannot be achieved by the text instructions alone. Experimental results show that our methods outperform existing approaches in preserving both music content consistency and editing fidelity. User studies further validate that our methods achieve superior music editing quality.

Evaluating a graphic design involves assessing it from multiple facets like alignment, composition, aesthetics and color choices. Holistic evaluation would involve aggregating feedback from individual expert reviewers. Towards this, we propose an Agentic Design Review System (Agentic-DRS), where multiple agents collaboratively analyze a design, orchestrated by a meta-agent. A novel in-context exemplar selection approach based on graph matching and a unique prompt expansion method plays central role towards making each agent design aware. In order to evaluate this framework, we propose DRS-BENCH. Thorough experimental evaluation against state-of-the-art baselines adapted to the problem setup, backed by critical ablations, demonstrates efficacy of Agentic-DRS in evaluating designs and generating actionable feedback.

Spiking neural networks (SNNs) promise highly energy-efficient computing, but their adoption is hindered by a critical scarcity of event-stream data. This work introduces I2E, an algorithmic framework that resolves this bottleneck by converting static images into high-fidelity event streams. By simulating microsaccadic eye movements with a highly parallelized convolution, I2E achieves a conversion speed over 300x faster than prior methods, uniquely enabling on-the-fly data augmentation for SNN training. The framework's effectiveness is demonstrated on large-scale benchmarks. An SNN trained on the generated I2E-ImageNet dataset achieves a state-of-the-art accuracy of 60.50%. Critically, this work establishes a powerful sim-to-real paradigm where pre-training on synthetic I2E data and fine-tuning on the real-world CIFAR10-DVS dataset yields an unprecedented accuracy of 92.5%. This result validates that synthetic event data can serve as a high-fidelity proxy for real sensor data, bridging a long-standing gap in neuromorphic engineering. By providing a scalable solution to the data problem, I2E offers a foundational toolkit for developing high-performance neuromorphic systems. The open-source algorithm and all generated datasets are provided to accelerate research in the field.

Spiking Neural Networks (SNNs), with their brain-inspired spatiotemporal dynamics and spike-driven computation, have emerged as promising energy-efficient alternatives to Artificial Neural Networks (ANNs). However, existing SNNs typically replicate inputs directly or aggregate them into frames at fixed intervals. Such strategies lead to neurons receiving nearly identical stimuli across time steps, severely limiting the model's expressive power—particularly in complex tasks like object detection. In this work, we propose the Temporal Dynamics Enhancer (TDE) to strengthen SNNs' capacity for temporal information modeling. TDE consists of two modules: a Spiking Encoder (SE) that generates diverse input stimuli across time steps, and an Attention Gating Module (AGM) that guides the SE generation based on inter-temporal dependencies. Moreover, to eliminate the high-energy multiplication operations introduced by the AGM, we propose a Spike-Driven Attention (SDA) to reduce attention-related energy consumption. Extensive experiments demonstrate that TDE can be seamlessly integrated into existing SNN-based detectors and consistently outperforms state-of-the-art methods, achieving mAP@50-95 scores of 57.7% on the static PASCAL VOC dataset and 47.6% on the neuromorphic EvDET200K dataset. In terms of energy consumption, the SDA consumes only 0.240× the energy of conventional attention modules.

Humans constantly generate a diverse range of tasks guided by internal motivations. While generative agents powered by large language models (LLMs) aim to simulate this complex behavior, it remains uncertain whether they operate on similar cognitive principles. To address this, we conducted a task-generation experiment comparing human responses with those of an LLM agent (GPT-4o). We find that human task generation is consistently influenced by psychological drivers, including personal values (e.g., Openness to Change) and cognitive style. Even when these psychological drivers are explicitly provided to the LLM, it fails to reflect the corresponding behavioral patterns. They produce tasks that are markedly less social, less physical, and thematically biased toward abstraction. Interestingly, while the LLM's tasks were perceived as more fun and novel, this highlights a disconnect between its linguistic proficiency and its capacity to generate human-like, embodied goals. We conclude that there is a core gap between the value-driven, embodied nature of human cognition and the statistical patterns of LLMs, highlighting the necessity of incorporating intrinsic motivation and physical grounding into the design of more human-aligned agents.

Multimodal sentiment analysis remains a challenging task due to the inherent heterogeneity across modalities. Such heterogeneity often manifests as asynchronous signals, imbalanced information between modalities, and interference from task-irrelevant noise, hindering the learning of robust and accurate sentiment representations. To address these issues, we propose a factorized multimodal fusion framework that first disentangles each modality into shared and unique representations, and then suppresses task-irrelevant noise within both to retain only sentiment-critical representations. This fine-grained decomposition improves representation quality by reducing redundancy, prompting cross-modal complementarity, and isolating task-relevant sentiment cues. Rather than manipulating the feature space directly, we adopt a mutual information–based optimization strategy to guide the factorization process in a more stable and principled manner. To further support feature extraction and long-term temporal modeling, we introduce two auxiliary modules: a Mixture of Q-Formers, placed before factorization, which precedes the factorization and uses learnable queries to extract fine-grained affective features from multiple modalities, and a Dynamic Contrastive Queue, placed after factorization, which stores latest high-level representations for contrastive learning, enabling the model to capture long-range discriminative patterns and improve class-level separability. Extensive experiments on multiple public datasets demonstrate that our method consistently outperforms existing approaches, validating the effecti veness and robustness of the proposed framework.

Spiking Neural Networks (SNNs) become popular due to excellent energy efficiency, yet facing challenges for effective model training. Recent works improve this by introducing knowledge distillation (KD) techniques, with the pre-trained artificial neural networks (ANNs) used as teachers and the target SNNs as students. This is commonly accomplished through a straightforward element-wise alignment of intermediate features and prediction logits from ANNs and SNNs, often neglecting the intrinsic differences between their architectures. Specifically, ANN's outputs exhibit a continuous distribution, whereas SNN's outputs are characterized by sparsity and discreteness. To mitigate this issue, we introduce two innovative KD strategies. Firstly, we propose the Saliency-scaled Activation Map Distillation (SAMD), which aligns the spike activation map of the student SNN with the class-aware activation map of the teacher ANN. Rather than performing KD directly on the raw features of ANN and SNN, our SAMD directs the student to learn from saliency activation maps that exhibit greater semantic and distribution consistency. Additionally, we propose a Noise-smoothed Logits Distillation (NLD), which utilizes Gaussian noise to smooth the sparse logits of student SNN, facilitating the alignment with continuous logits from teacher ANN. Extensive experiments on multiple datasets demonstrate the effectiveness of our methods.

Spiking Neural Networks (SNNs) are emerging as a promising energy-efficient alternative to Artificial Neural Networks (ANNs) due to their event-driven computation paradigm. However, recent advances toward large-scale high-performance SNNs inevitably lead to substantial memory and computational overhead. While quantization offers a potential way, many quantization approaches fail to deliver verifiable efficiency gains on resource-constrained hardware platforms. In this paper, we propose a lightweight and hardware-friendly SNN, termed HardF-SNN. Specifically, we first build a baseline model using shared-scale quantization and BN folding to simulate integer-only inference, as this has not been thoroughly discussed in prior SNN works. Then, through empirical and theoretical analysis, we identify that the baseline suffers from accuracy degradation and may cause training failure. To mitigate these issues, we propose proportional shared-scale quantization for enhanced dynamic range and integer-only BN using bit-shifting to stabilize training. Extensive experiments show that HardF-SNN achieves an optimal balance between performance and efficiency with excellent hardware compatibility. To demonstrate its effectiveness on resource-limited platforms, HardF-SNN is deployed on a dedicated FPGA-based hardware accelerator. Evaluation results indicate that our implementation achieves significant performance improvements over several existing hardware accelerators.