United States

A Metacognitive Framework for Multi-Table Question Answering
Processing structured data is critical in finance, healthcare, and science. While single-table question answering has advanced, multi-table QA remains challenging due to schema understanding, cross-table reasoning, and complex natural language queries. We propose M2TQA , a novel framework inspired by human cognitive and metacognitive mechanisms. M2TQA integrates metadata extraction, query decomposition, and a metacognitive module to enable interpretable, robust solutions for MTQA. It dynamically simulates human-like reasoning through feedback loops, bridging gaps between natural language understanding and structured data processing. Experiments on four benchmarks show M2TQA outperforms baselines by 94.54% and 33.24% in F1 scores. This work advances MTQA and highlights metacognition’s role in AI, fostering interdisciplinary connections between cognitive science and artificial intelligence.

CogSci 2025

M2TQA:A Metacognitive Framework for Multi-Table Question Answering

cognitive architectures

agent-based modeling

cognitive neuroscience

problem solving

artificial intelligence

It is my great pleasure and honor to welcome you to the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), held in San Diego, California, USA, July 2–7, 2026. ACL 2026 continues our fields tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.

It is exciting to see the many developments in the field reflected at our conference, such as the advancement of large language models (born out of our community) into many application directions. This is not without challenges, partly due to unsolved problems, and partly due to misuse of the technology. All this is reflected well at ACL 2026. The conference also reflects that the field is not only about advancing technology, but also the science of fundamentally understanding language and how it is use by many different speaker communities. <br>
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) <br><br><br>
<html><button style="display: inline-flex; align-items: center; justify-content: center; white-space: nowrap; border-radius: 9999px; font-weight: bold; background: #7c3aed; color: white; font-family: 'Space Grotesk', sans-serif; height: 40px; font-size: 16px; padding: 0 20px; border: none; cursor: pointer" onclick="window.open('https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing')">Conference Handbook</button></html>
<br><br><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 ACL 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://2026.aclweb.org/registration/) first.

**Online Registration Form**: https://acl.swoogo.com/acl2026

Registration Required

ACL 2026

The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) will take place in San Diego, California, from July 2 - 7, 2026.

poster

### Welcome to CogSci Conference 2025!

The 47th Annual Meeting of the Cognitive Science Society was a hybrid meeting held in San Francisco. 

<div style="position:relative;padding-top:0;width:900px;height:500px;"><iframe style="position:absolute;border:none;width:100%;height:100%;left:0;top:0;" src="https://online.fliphtml5.com/ebtyf/amvr/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe></div>

#### About

The Cognitive Science Society brings together researchers from around the world who hold a common goal: understanding the nature of the human mind. The mission of the Society is to promote Cognitive Science as a discipline, and to foster scientific interchange among researchers in various areas of study, including Artificial Intelligence, Linguistics, Anthropology, Psychology, Neuroscience, Philosophy, and Education.

The Society is a non-profit professional organization and its activities include sponsoring an annual conference and publishing the journals Cognitive Science and TopiCS.

#### Our History 

* **Society Creation**<br>
The Society was incorporated as a 501(c)(3) non-profit professional organization in Massachusetts in 1979. The organizing committee included Roger Schank, Allan Collins, Donald Norman, and a number of other scholars from psychology, linguistics, computer science, and philosophy. 
<br><br>
* **Conference Creation**<br>
The first conference on cognitive science was held at La Jolla, California in August, 1979, and has occurred annually since then. The proceedings of each conference are published, and those from most years are available through Lawrence Erlbaum Associates, Inc. The annual proceedings of the Cognitive Science Conference represent a major source of information on new work and new ideas in the scientific study of thinking. In 1990, the Society, with help from an anonymous donor, established the David Marr Prize for the best student paper at each annual meeting.
<br><br>
* **Journal Creation**<br>
The Journal, Cognitive Science, began publication in 1976, and is now published by Wiley-Blackwell. The Executive Editor is currently Richard P. Cooper of Birkbeck, University of London, and there are 18 Associate Editors and a 30-member editorial board. It serves as the premier outlet for research reports that intersect two or more disciplines. Copyrights for articles published in the journal are held by the Society. The Governing Board of the Cognitive Science Society voted in late 2006 to found a new journal, Topics in Cognitive Science (topiCS). The Editor in Chief is Wayne Gray, Cognitive Science Department, Rensselaer Polytechnic Institute. The journal seeks to fill a niche not occupied by Cognitive Science Journal or other cognitive science journals. Membership in the Society includes a subscription to Cognitive Science and TopiCS. Copyrights for articles published in the journal are held by the Society.
<br><br>

#### Code of Conduct

By attending the CogSci 2025 Conference, you are required to adhere to the society’s **[Code of Conduct](https://drive.google.com/file/d/1ChPuihLy6jE_BWqfO7J2KKgX35JW2zsM/view?usp=sharing)**.
<br><br>


You need to log in with the email address you registered with. 

Login credentials were sent to you from Underline -  subject line "Welcome to the CogSci 2025 Conference". Please be sure to check your spam/promotional inbox  if you do not see an email confirmation right away.





Please log in to join this event.

To access the site, please register [**here**](https://cognitivesciencesociety.org/registration/).

If you are registered and feel like you are seeing this message by mistake, please make sure you are logged in with the same email that you registered with. 

Please register!

The 47th Annual Meeting of the Cognitive Science Society presents the latest research across cognitive science and highlights the theme of Cognition in Context.

Emotion recognition based on electroencephalography (EEG) plays a significant role in brain-computer interface (BCI) applications. However, individual differences often hinder the generalization of emotion recognition methods to unknown subjects. To address this, we propose an unsupervised domain adaptive model with joint multi-source domain and two-step alignment strategy (JMS2A). The alignment strategy consists of two steps: (1) To capture the structured information from source domain, we combine multiple source domains into a mixed source domain. Simultaneously, a single source domain and the target domain are combined to form a pseudo-target domain, which is then indirectly aligned with the mixed source domain; (2) To extract latent class information from the target domain, we extend supervised contrastive learning to enable direct alignment between source and target domain. We evaluated JMS2A on the SEED and SEED-IV datasets, achieving accuracies of 95.30% and 86.55%, respectively. Experimental results demonstrate that our approach outperforms state-of-the-art methods. The source code is available at https://github.com/cccyangt/JMS2A.

JMS2A: Joint Multi-source Domain and Two-step Alignment Strategy for Cross-subject EEG Emotion Recognition

The Social Intuitionist Model emphasizes the primary role of moral intuitions, rather than analytical moral reasoning, in forming moral judgments. In line with this model, moral emotions, rather than reasoning, are strongly associated with moral actions. This study explored the impact of emotionally evocative film excerpts, specifically those featuring moral dilemmas within varying contexts of social distance, on moral judgment processes. We invited 88 college students to view excerpts from foreign movies and their Chinese remakes under three music conditions: no music, positive music, and negative music. Participants evaluated the characters' actions based on three moral dimensions: Harm, Fairness, and Authority, while also rating their emotional responses. The results indicated that positive music during moral dilemmas elicited strong negative emotional responses. Additionally, viewers judged the protagonist's actions more harshly regarding fairness and placed greater emphasis on social order and authority in Chinese compared to foreign movie excerpts, regardless of the music's emotional valence.

Emotional Resonance in Film: Disentangling the Effects of Music on Moral Judgments in Dilemmas

Automatic seizure detection leveraging artificial intelligence has gained widespread attention. However, existing research has predominantly focused on scenarios with patient-specific and fixed-time lengths, with the practical clinical applications across non-specific patients and variable time lengths remaining underexplored. To address this gap, we introduce a novel method named Electroencephalogram Channel-wise Normalization (ECNorm), designed to thoroughly explore the physical significance and data distribution characteristics of different EEG channels to minimize inter-patient variability. We applied ECNorm to a two-layer LSTM model to facilitate cross-patient adaptive-length epilepsy diagnosis. Ablation studies demonstrate that ECNorm significantly enhances the performance of simple architectures like the two-layer LSTM when compared to batch normalization and layer normalization. Leave-one-out experiments on the public CHB-MIT dataset verify that our approach surpasses existing studies across segments of varying lengths (1 and 100 seconds), establishing a new benchmark for patient-independent automated epilepsy diagnosis.

Empowering Cross-Patient Adaptive-Length Epilepsy Diagnosis with ECNorm: A Channel-wise Approach

This study investigates whether large language models (LLMs) possess human-like syntactic competence by examining their handling of English center-embedded structures. Two experiments were conducted: one collected acceptability judgments from native speakers, confirming sensitivity to syntactic constraints; the other measured surprisal values from GPT-2 and Gemma 2 on grammatical versus ungrammatical center-embedded sentences. While both models distinguished between the two types probabilistically, they failed to replicate categorical human judgments. The results suggest that LLMs conflate low-frequency constructions with ungrammaticality, reflecting limitations in hierarchical syntactic understanding. It is argued that genuine linguistic competence requires more than statistical pattern recognition and advocate for integrating formal syntactic theory into model development. This work contributes to the ongoing dialogue between generative linguistics and AI, highlighting key distinctions between human cognition and current LLMs.

Evaluating the Linguistic Competence of Large Language Models: Experimental Evidence from Center-embedding Structures

In some cases, cognitive function can change owing to external factors without physiological effects. This is considered to be influenced by emotional arousal caused by schema activity. This study analyzed the relationships between emotional arousal and task performance to determine whether schema can lead to emotional arousal and enhance task performance. Emotional arousal was assessed based on pupil dilation using eye tracking, and task performance was measured based on reaction times to the stimuli. No main effect of condition on reaction time was observed. However, the experimental results showed that the larger the pupil dilation, the shorter the reaction time, but only when schema activation stimuli were present. These findings suggest that schema-induced emotional arousal may improve performance without the intake of substances with physiological effects. This may also offer insight into mechanisms underlying the placebo effect, though further investigation is required.

Schema-Induced Emotional Arousal Enhances Task Performance:  A Pupillometric Investigation of Top-Down Cognitive Influence

Recommender systems personalize content delivery based on user’s interaction history. However, not all clicks result from deliberate decisions—many arise from intuitive reactions. Inspired by the dual process theory, we argue that intuitive clicks are primarily driven by System 1, reacting to superficial cues, while analytical clicks involve deeper processing by System 2, considering the semantic meaning and long-term preference. However, existing models overlook these cognitive mechanisms. To address this, we propose DualRec, a novel recommendation method that models both intuitive and analytical behaviors. DualRec encodes items using language models, leveraging shallow layers for superficial understanding (System 1) and deep layers for semantic comprehension (System 2). It employs Transformer-based encoders with two attention mechanisms to capture intuitive “fast” and analytical “slow” click patterns. A learnable fusion layer balances these behaviors. Extensive experiments demonstrate that DualRec outperforms existing methods and highlights the importance of integrating both cognitive processes in recommendations.

Clicking, Fast and Slow: Towards Intuitive and Analytical Behaviors Modeling for Recommender Systems

Cross-Domain Sequential Recommendation (CDSR) predicts user behavior by leveraging historical interactions across multiple domains, focusing on modeling cross-domain preferences through intra- and inter-sequence item relationships. Inspired by human cognitive processes, we propose Hierarchical Attention Fusion of Visual and Textual Representations (HAF-VT), a novel approach integrating visual and textual data to enhance cognitive modeling. Using the frozen CLIP model, we generate image and text embeddings, enriching item representations with multimodal data. A hierarchical attention mechanism jointly learns single-domain and cross-domain preferences, mimicking human information integration. Evaluated on four e-commerce datasets, HAF-VT outperforms existing methods in capturing cross-domain user interests, bridging cognitive principles with computational models and highlighting the role of multimodal data in sequential decision-making.

Cognitive-Inspired Hierarchical Attention Fusion With Visual and Textual for Cross-Domain Sequential Recommendation

Brain-to-Image reconstruction aims to recover visual stimuli perceived by humans from brain activity. However, the reconstructed visual stimuli often missing details and semantic inconsistencies, which may be attributed to insufficient semantic information. To address this issue, we propose an approach named Fine-grained Brain-to-Image reconstruction (FgB2I), which employs fine-grained text as bridge to improve image reconstruction. FgB2I comprises three key stages: detail enhancement, decoding fine-grained text descriptions, and text-bridged brain-to-image reconstruction. In the detail-enhancement stage, we leverage large vision–language models to generate fine-grained captions for visual stimuli and experimentally validate its importance. We propose three reward metrics (object accuracy, text-image semantic similarity, and image-image semantic similarity) to guide the language model in decoding fine-grained text descriptions from fMRI signals. The fine-grained text descriptions can be integrated into existing reconstruction methods to achieve fine-grained Brain-to-Image reconstruction.

Improving Brain-to-Image Reconstruction via Fine-Grained Text Bridging

Electroencephalogram (EEG) has become an important indicator reflecting emotions. Due to its natural graph structure characteristics, it has made significant progress in the emotional recognition using graph convolutional networks (GCN). However, existing methods face limitations: (1) insufficient integration of psychological prior knowledge, limiting the utilization of brain activity patterns, and (2) simplistic node relationship construction, neglecting the universality and functional connectivity of brain regions. Therefore, we propose a dual-path parallel graph convolutional network (DP-GCN). The first path leverages psychological prior knowledge to segment electrodes into brain regions and employs an attention mechanism to integrate features. The second approach employs a data-driven method, using a sparse stacked autoencoder to reconstruct brain region features, while a learnable, input-independent adjacency matrix captures EEG patterns associated with emotions. Finally, a cross-attention mechanism integrates features from both paths. DP-GCN has been evaluated on public dataset, achieving an accuracy of 82.69%±4.16%, demonstrating its competitive performance.

Dual-Path Parallel Graph Convolution Combining Brain Region Partitioning and Data-Driven Learning for EEG Emotion Recognition

Emotion recognition is crucial for applications like human-computer interaction, safe driving, and remote education, with EEG-based methods providing more authentic insights than facial or speech-based approaches. In this paper, we propose an Adaptive Multiview Fusion Transformer (AMFT) that effectively fuses Differential Entropy (DE) and Power Spectral Density (PSD) features in EEG signals. AMFT consists of three main components—a Multi-Perspective Embedder (MPE), a Dual Cross Attention Module (DCAM), and a ClsTransformerEncoder (CTE)—which use multiview projection and iterative cross-attention to combine DE and PSD features. Experiments on the SEED-VII dataset show that AMFT achieves higher accuracy and F1 scores in both multi-class and binary-class emotion recognition tasks, with improved stability compared to existing methods. Ablation studies confirm the key role of the multiview embedder and cross-attention module in boosting performance, offering new insights for EEG-based emotion recognition and biomedical signal analysis. Our code is available at https://github.com/sizhiyier/AMFT.

Downloads

Next from CogSci 2025

JMS2A: Joint Multi-source Domain and Two-step Alignment Strategy for Cross-subject EEG Emotion Recognition

Similar lecture

EASE: Entity-Aware Sub-table Generation for Real-world Multi-table QA