More Like this

1. A Three-Tier LLM Framework for Forecasting Student Engagement from Qualitative Longitudinal Data

   https://doi.org/10.18653/v1/2025.conll-1.22  

   Hayat, Ahatsham; Martinez, Helen; Khan, Bilal; Hasan, Mohammad Rashedul (January 2025, Association for Computational Linguistics)

   Not AvailableForecasting nuanced shifts in student engagement from longitudinal experiential (LE) data—multi-modal, qualitative trajectories of academic experiences over time—remains challenging due to high dimensionality and missingness. We propose a natural language processing (NLP)-driven framework using large language models (LLMs) to forecast binary engagement levels across four dimensions: Lecture Engagement Disposition, Academic Self-Efficacy, Performance Self-Evaluation, and Academic Identity and Value Perception. Evaluated on 960 trajectories from 96 first-year STEM students, our three-tier approach—LLM-informed imputation to generate textual descriptors for missing-not-at-random (MNAR) patterns, zero-shot feature selection via ensemble voting, and fine-tuned LLMs—processes textual non-cognitive responses. LLMs substantially outperform numeric baselines (e.g., Random Forest, LSTM) by capturing contextual nuances in student responses. Encoder-only LLMs surpass decoder-only variants, highlighting architectural strengths for sparse, qualitative LE data. Our framework advances NLP solutions for modeling student engagement from complex LE data, excelling where traditional methods struggle. 

   more » « less
2. MUSE: A Multimodal, Generative, and Symbolic Framework for Human Experience Modeling

   https://doi.org/10.1145/3716553.3750735  

   Hasan, Mohammad Rashedul (October 2025, ACM)

   Realizing truly capable personal AI for health and education requires effectively modeling complex longitudinal experiential (LE) data. Unlike standard datasets, LE data from human experience is inherently multifaceted, dynamic, and contextual. Current AI approaches struggle with this complexity due to three critical gaps rooted in differences from human cognition: insufficient multimodal processing, lack of generative perception, and inadequate symbolic order contextualization. Drawing upon interdisciplinary insights, our blue sky vision, the MUSE (Multimodal, generative, and Symbolic framEwork), proposes to bridge these gaps. By integrating multimodal representations addressing LE data's structure, generative simulation capturing "what-if" dynamics, and symbolic order grounding for context, MUSE aims for a profound, contextually aware understanding of individual experience, essential for robust inference and enabling advanced personal AI applications. 

   more » « less
3. DreamRunner: Fine-Grained Compositional Story-to-Video Generation with Retrieval-Augmented Motion Adaptation

   https://doi.org/10.1609/aaai.v40i13.38022  

   Wang, Zun; Li, Jialu; Lin, Han; Yoon, Jaehong; Bansal, Mohit (March 2026, Proceedings of the AAAI Conference on Artificial Intelligence)

   Storytelling video generation (SVG) aims to produce coherent and visually rich multi-scene videos that follow a structured narrative. Existing methods primarily employ LLM for high-level planning to decompose a story into scene-level descriptions, which are then independently generated and stitched together. However, these approaches struggle with generating high-quality videos aligned with the complex single-scene description, as visualizing such complex description involves coherent composition of multiple objects/events, complex motion synthesis and character customization with sequential motions. To address these challenges, we propose DREAMRUNNER, a novel story-to-video generation method: First, we structure the input script using a large language model (LLM) to facilitate both coarse-grained scene planning as well as fine-grained object-level layout planning. Next, DREAMRUNNER presents retrieval-augmented test-time adaptation to capture target motion priors for objects in each scene, supporting diverse motion customization based on retrieved videos, thus facilitating the generation of new videos with complex, scripted motions. Lastly, we propose a novel spatial-temporal region-based 3D attention and prior injection module SR3AI for fine-grained object-motion binding and frame-by-frame spatial-temporal semantic control. We compare DREAMRUNNER with various SVG baselines, demonstrating state-of-the-art performance in character consistency, text alignment, and smooth transitions. Additionally, DREAMRUNNER exhibits strong fine-grained condition-following ability in compositional text-to-video generation, significantly outperforming baselines on T2V-ComBench. Finally, we demonstrate DREAMRUNNER’s ability to generate multi-character interactions with qualitative examples. 

   more » « less
4. Hybrid Sequential Pattern Mining and Machine Learning for Context-Aware Event Prediction in IoT

   https://doi.org/10.1109/CCWC67433.2026.11393795  

   Goyal, Pooja; Helmholz, Bradley; Tran, Trinh; Arnob, Farhan Rahman; Bryce, Renee (January 2026, IEEE)

   Context-aware event prediction is a critical challenge in Internet of Things (IoT) systems, where accurate forecasting enables adaptive and personalized applications. Existing approaches often rely on either sequential pattern mining or machine learning models in isolation, limiting their ability to capture complex temporal dependencies. In this work, we introduce a hybrid framework that integrates sequential pattern mining-via POERMH (Partially-Ordered Episode Rule Mining with Head support) and PrefixSpan-with discriminative learning models to enhance context-event prediction on mobile IoT devices. Our key contribution lies in leveraging mined sequential structures as feature representations for classifiers, combined with optimized time-interval segmentation to preserve temporal dynamics. Empirical evaluation across real-world IoT datasets demonstrates that this integration significantly outperforms standalone baselines, with Support Vector Machines achieving an F1-score of 67.08 %, representing a substantial improvement over traditional sequence-only methods. These findings highlight the importance of coupling symbolic pattern discovery with statistical learning for robust context modeling, offering a scalable approach for next-generation IoT applications. 

   more » « less
5. Modeling Latent Human Factors in Construction Reports through ReAct-Based Reasoning

   Ray, Unmesa; Park, JeeWoong; Anderson, Brian; Kim; Namgyun (November 2025, Korea Institute of Construction Engineering and Management)

   Unstructured construction accident narratives often contain latent behavioral signals, such as fatigue, overconfidence, or rushing. These allow an enhanced understanding of accident context but are rarely captured through traditional extraction methods. This paper presents a modular ReAct (Reasoning + Acting) agent framework designed to infer contextually relevant behavioral contributor in each narrative. The ReAct paradigm enables the agent to engage in stepby-step reasoning, invoke specialized tools, and iteratively refine its understanding through Thought-Action-Observation loops. This structure addresses key limitations of static Large Language Model (LLM) outputs by introducing transparent, modular, and adaptable reasoning. Our contribution lies in operationalizing this framework for behavioral insight generation. We achive this by developing and applying a behavior prioritization module to a realworld dataset of construction accident reports. This modules enables the agent to produce structured outputs including selected behaviors, confidence levels, and justifications, which in turn facilitates a behavior-aware interpretation of construction accidents. This is accomplished without requiring supervision or fine-tuning of Foundation Models. This work highlights how ReAct-style agents can deepen narrative understanding and offer a foundation for explainable behavioral analytics in safety-critical domains. 

   more » « less