reasoning from minimal examples remains a core unsolved problem for frontier foundation models such as GPT-5 and Grok 4. These models still fail to infer structured transformation rules from a handful of examples, which is a key hallmark of human intelligence. The Abstraction and Reasoning Corpus for Artificial General Intelligence (ARC-AGI) provides a rigorous testbed for this capability, demanding conceptual rule induction and transfer to novel tasks. Most existing methods treat ARC-AGI as a purely textual reasoning task, overlooking the fact that humans rely heavily on visual abstraction when solving such puzzles. However, our pilot experiments reveal a paradox: naively rendering ARC-AGI grids as images degrades performance due to imprecise rule execution. This leads to our central hypothesis that vision and language possess complementary strengths across distinct reasoning stages: vision supports global pattern abstraction and verification, whereas language specializes in symbolic rule formulation and precise execution. Building on this insight, we introduce two synergistic strategies: (1) Vision-Language Synergy Reasoning (VLSR), which decomposes ARC-AGI into modality-aligned subtasks; and (2) Modality-Switch Self-Correction (MSSC), which leverages vision to verify text-based reasoning for intrinsic error correction. Extensive experiments demonstrate that our approach yields up to a 4.33% improvement over text-only baselines across diverse flagship models and multiple ARC-AGI tasks. Our findings suggest that unifying visual abstraction with linguistic reasoning is a crucial step toward achieving general-izable, human-like intelligence in future foundation models. Source code is released at https://github.com/

OpenAI's o3-preview reasoning model exceeded human accuracy on the ARC-AGI benchmark, but does that mean state-of-the-art models recognize and reason with the abstractions that the task creators intended? We investigate models' abstraction abilities on ConceptARC. We evaluate models under settings that vary the input modality (textual vs. visual), whether the model is permitted to use external Python tools, and, for reasoning models, the amount of reasoning effort. In addition to measuring output accuracy, we perform fine-grained evaluation of the natural-language rules that models generate to explain their solutions. This dual evaluation lets us assess whether models solve tasks using the abstractions ConceptARC was designed to elicit, rather than relying on surface-level patterns. Our results show that, while some models using text-based representations match human output accuracy, the best models' rules are often based on surface-level ``shortcuts'' and capture intended abstractions far less often than humans. Thus their capabilities for general abstract reasoning may be overestimated by evaluations based on accuracy alone. In the visual modality, AI models' output accuracy drops sharply, yet our rule-level analysis reveals that models might be underestimated, as they still exhibit a substantial share of rules that capture intended abstractions, but are often unable to correctly apply these rules. In short, our results show that models still lag humans in abstract reasoning, and that using accuracy alone to evaluate abstract reasoning on ARC-like tasks may overestimate abstract-reasoning capabilities in textual modalities and underestimate it in visual modalities. We believe that our evaluation framework offers a more faithful picture of multimodal models' abstract reasoning abilities and a more principled way to track progress toward human-like, abstraction-centered intelligence.

Retrieval-Augmented Generation (RAG) has become a core paradigm in document question answering tasks. However, existing methods have limitations when dealing with multimodal documents: one category of methods relies on layout analysis and text extraction, which can only utilize explicit text information and struggle to capture images or unstructured content; the other category treats document segmentation as visual input and directly passes it to visual language models (VLMs) for processing, yet it ignores the semantic advantages of text, leading to suboptimal retrieval and generation results. To address these research gaps, we propose the Co-Modality-based RAG (CMRAG) framework, which can simultaneously leverage texts and images for more accurate retrieval and generation. Our framework includes two key components: (1) a Unified Encoding Model (UEM) that projects queries, parsed text, and images into a shared embedding space via triplet-based training, and (2) a Unified Co-Modality-informed Retrieval (UCMR) method that statistically normalizes similarity scores to effectively fuse cross-modal signals. To support research in this direction, we further construct and release a large-scale triplet dataset of (query, text, image) examples. Experiments demonstrate that our proposed framework consistently outperforms single-modality--based RAG in multiple visual document question-answering (VDQA) benchmarks. The findings of this paper show that integrating co-modality information into the RAG framework in a unified manner is an effective approach to improving the performance of complex VDQA systems.

The scarcity of large-scale and high-quality electronic health records (EHRs) remains a major bottleneck in biomedical research, especially as large foundation models become increasingly data-hungry. Synthesizing substantial volumes of de-identified and high-fidelity data from existing datasets has emerged as a promising solution. However, existing methods suffer from a series of limitations: they struggle to model the intrinsic properties of heterogeneous multimodal EHR data (e.g., continuous, discrete, and textual modalities), capture the complex dependencies among them, and robustly handle pervasive data incompleteness. These challenges are particularly acute in Traditional Chinese Medicine (TCM). To this end, we propose TCDiff (Triplex Cascaded Diffusion Network), a novel EHR generation framework that cascades three diffusion networks to learn the features of real-world EHR data, formatting a multi-stage generative process: Reference Modalities Diffusion, Cross-Modal Bridging, and Target Modality Diffusion. Furthermore, to validate our proposed framework, besides two public datasets, we also construct and introduce TCM-SZ1, a novel multimodal EHR dataset for benchmarking. Experimental results show that TCDiff consistently outperforms state-of-the-art baselines by an average of 10% in data fidelity under various missing rate, while maintaining competitive privacy guarantees. This highlights the effectiveness, robustness, and generalizability of our approach in real-world healthcare scenarios.

Multimodal emotion recognition in conversation (MER) aims to accurately identify emotions in conversational utterances by integrating multimodal information. Previous methods usually treat multimodal information as equal quality and employ symmetric architectures to conduct multimodal fusion. However, in reality, the quality of different modalities usually varies considerably, and utilizing a symmetric architecture is difficult to accurately recognize conversational emotions when dealing with uneven modal information. Furthermore, fusing multi-modality information in a single granularity may fail to adequately integrate modal information, exacerbating the inaccuracy in emotion recognition. In this paper, we propose a novel Cross-Modality Augmented Transformer with Hierarchical Variational Distillation, called CMATH, which consists of two major components, i.e., Multimodal Interaction Fusion and Hierarchical Variational Distillation. The former is comprised of two submodules, including Modality Reconstruction and Cross-Modality Augmented Transformer (CMA-Transformer), where Modality Reconstruction focuses on obtaining high-quality compressed representation of each modality, and CMA-Transformer adopts an asymmetric fusion strategy which treats one modality as the central modality and takes others as auxiliary modalities. The latter first designs a variational fusion network to fuse the fine-grained representations learned by CMA- Transformer into a coarse-grained representations. Then, it introduces a hierarchical distillation framework to maintain the consistency between modality representations with different granularities. Experiments on the IEMOCAP and MELD datasets demonstrate that our proposed model outperforms previous state-of-the-art baselines. Implementation codes can be available at https://github.com/ cjw-MER/CMATH.

Depression has proven to be a significant public health issue, profoundly affecting the psychological well-being of individuals. If it remains undiagnosed, depression can lead to severe health issues, which can manifest physically and even lead to suicide. Generally, Diagnosing depression or any other mental disorder involves conducting semi-structured interviews alongside supplementary questionnaires, including variants of the Patient Health Questionnaire (PHQ) by Clinicians and mental health professionals. This approach places significant reliance on the experience and judgment of trained physicians, making the diagnosis susceptible to personal biases. Given that the underlying mechanisms causing depression are still being actively researched, physicians often face challenges in diagnosing and treating the condition, particularly in its early stages of clinical presentation. Recently, significant strides have been made in Artificial neural computing to solve problems involving text, image, and speech in various domains. Our analysis has aimed to leverage these state-of-the-art (SOTA) models in our experiments to achieve optimal outcomes leveraging multiple modalities. The experiments were performed on the Extended Distress Analysis Interview Corpus Wizard of Oz dataset (E-DAIC) corpus presented in the Audio/Visual Emotion Challenge (AVEC) 2019 Challenge. The proposed solutions demonstrate better results achieved by Proprietary and Open-source Large Language Models (LLMs), which achieved a Root Mean Square Error (RMSE) score of 3.98 on Textual Modality, beating the AVEC 2019 challenge baseline results and current SOTA regression analysis architectures. Additionally, the proposed solution achieved an accuracy of 71.43% in the classification task. The paper also includes a novel audio-visual multi-modal network that predicts PHQ-8 scores with an RMSE of 6.51.

This paper presents our winning submission to Subtask 2 of SemEval 2024 Task 3 on multimodal emotion cause analysis in conversations. We propose a novel Multimodal Emotion Recognition and Multimodal Emotion Cause Extraction (MER-MCE) framework that integrates text, audio, and visual modalities using specialized emotion encoders. Our approach sets itself apart from top-performing teams by leveraging modality-specific features for enhanced emotion understanding and causality inference. Experimental evaluation demonstrates the advantages of our multimodal approach, with our submission achieving a competitive weighted F1 score of 0.3435, ranking third with a margin of only 0.0339 behind the 1st team and 0.0025 behind the 2nd team. Project: https://github.com/MIPS-COLT/MER-MCE.git

Sign language translation (SLT) aims to convert continuous sign language videos into textual sentences. As a typical multi-modal task, there exists an inherent modality gap between sign language videos and spoken language text, which makes the cross-modal alignment between visual and textual modalities crucial. However, previous studies tend to rely on an intermediate sign gloss representation to help alleviate the cross-modal problem thereby neglecting the alignment across modalities that may lead to compromised results. To address this issue, we propose a novel framework based on Conditional Variational autoencoder for SLT (CV-SLT) that facilitates direct and sufficient cross-modal alignment between sign language videos and spoken language text. Specifically, our CV-SLT consists of two paths with two Kullback-Leibler (KL) divergences to regularize the outputs of the encoder and decoder, respectively. In the prior path, the model solely relies on visual information to predict the target text; whereas in the posterior path, it simultaneously encodes visual information and textual knowledge to reconstruct the target text. The first KL divergence optimizes the conditional variational autoencoder and regularizes the encoder outputs, while the second KL divergence performs a self-distillation from the posterior path to the prior path, ensuring the consistency of decoder outputs. We further enhance the integration of textual information to the posterior path by employing a shared Attention Residual Gaussian Distribution (ARGD), which considers the textual information in the posterior path as a residual component relative to the prior path. Extensive experiments conducted on public datasets (PHOENIX14T and CSL-daily) demonstrate the effectiveness of our framework, achieving new state-of-the-art results while significantly alleviating the cross-modal representation discrepancy.

Multimodal sentiment analysis (MSA), which supposes to improve text-based sentiment analysis with associated acoustic and visual modalities, is an emerging research area due to its potential applications in Human-Computer Interaction (HCI). However, the existing researches observe that the acoustic and visual modalities contribute much less than the textual modality, termed as text-predominant. Under such circumstances, in this work, we emphasize making non-verbal cues matter for the MSA task. Firstly, from the resource perspective, we present the CH-SIMS v2.0 dataset, an extension and enhancement of the CH-SIMS. Compared with the original dataset, the CH-SIMS v2.0 doubles its size with another 2121 refined video segments with both unimodal and multimodal annotations and collects 10161 unlabelled raw video segments with rich acoustic and visual emotion-bearing context to highlight non-verbal cues for sentiment prediction. Secondly, from the model perspective, benefiting from the unimodal annotations and the unsupervised data in the CH-SIMS v2.0, the Acoustic Visual Mixup Consistent (AV-MC) framework is proposed. The designed modality mixup module can be regarded as an augmentation, which mixes the acoustic and visual modalities from different videos. Through drawing unobserved multimodal context along with the text, the model can learn to be aware of different non-verbal contexts for sentiment prediction. Our evaluations demonstrate that both CH-SIMS v2.0 and AV-MC framework enables further research for discovering emotion-bearing acoustic and visual cues and paves the path to interpretable end-to-end HCI applications for real-world scenarios.

Existing studies on multimodal sentiment analysis heavily rely on textual modality and unavoidably induce the spurious correlations between textual words and sentiment labels. This greatly hinders the model generalization ability. To address this problem, we define the task of out-of-distribution (OOD) multimodal sentiment analysis. This task aims to estimate and mitigate the bad effect of textual modality for strong OOD generalization. To this end, we embrace causal inference, which inspects the causal relationships via a causal graph. From the graph, we find that the spurious correlations are attributed to the direct effect of textual modality on the model prediction while the indirect one is more reliable by considering multimodal semantics. Inspired by this, we devise a model-agnostic counterfactual framework for multimodal sentiment analysis, which captures the direct effect of textual modality via an extra text model and estimates the indirect one by a multimodal model. During the inference, we first estimate the direct effect by the counterfactual inference, and then subtract it from the total effect of all modalities to obtain the indirect effect for reliable prediction. Extensive experiments show the superior effectiveness and generalization ability of our proposed framework.