Probing Scientific General Intelligence of LLMs with Scientist-Aligned Workflows

Despite advances in scientific AI, a coherent framework for Scientific General Intelligence (SGI)-the ability to autonomously conceive, investigate, and reason across scientific domains-remains lacking. We present an operational SGI definition grounded in the Practical Inquiry Model (PIM: Deliberation, Conception, Action, Perception) and operationalize it via four scientist-aligned tasks: deep research, idea generation, dry/wet experiments, and experimental reasoning. SGI-Bench comprises over 1,000 expert-curated, cross-disciplinary samples inspired by Science's 125 Big Questions, enabling systematic evaluation of state-of-the-art LLMs. Results reveal gaps: low exact match (10--20%) in deep research despite step-level alignment; ideas lacking feasibility and detail; high code executability but low execution result accuracy in dry experiments; low sequence fidelity in wet protocols; and persistent multimodal comparative-reasoning challenges. We further introduce Test-Time Reinforcement Learning (TTRL), which optimizes retrieval-augmented novelty rewards at inference, enhancing hypothesis novelty without reference answer. Together, our PIM-grounded definition, workflow-centric benchmark, and empirical insights establish a foundation for AI systems that genuinely participate in scientific discovery.

Scaffolding Metacognition in Programming Education: Understanding Student-AI Interactions and Design Implications

Generative AI tools such as ChatGPT now provide novice programmers with unprecedented access to instant, personalized support. While this holds clear promise, their influence on students' metacognitive processes remains underexplored. Existing work has largely focused on correctness and usability, with limited attention to whether and how students' use of AI assistants supports or bypasses key metacognitive processes. This study addresses that gap by analyzing student-AI interactions through a metacognitive lens in university-level programming courses. We examined more than 10,000 dialogue logs collected over three years, complemented by surveys of students and educators. Our analysis focused on how prompts and responses aligned with metacognitive phases and strategies. Synthesizing these findings across data sources, we distill design considerations for AI-powered coding assistants that aim to support rather than supplant metacognitive engagement. Our findings provide guidance for developing educational AI tools that strengthen students' learning processes in programming education.

MARS2 2025 Challenge on Multimodal Reasoning: Datasets, Methods, Results, Discussion, and Outlook

This paper reviews the MARS2 2025 Challenge on Multimodal Reasoning. We aim to bring together different approaches in multimodal machine learning and LLMs via a large benchmark. We hope it better allows researchers to follow the state-of-the-art in this very dynamic area. Meanwhile, a growing number of testbeds have boosted the evolution of general-purpose large language models. Thus, this year's MARS2 focuses on real-world and specialized scenarios to broaden the multimodal reasoning applications of MLLMs. Our organizing team released two tailored datasets Lens and AdsQA as test sets, which support general reasoning in 12 daily scenarios and domain-specific reasoning in advertisement videos, respectively. We evaluated 40+ baselines that include both generalist MLLMs and task-specific models, and opened up three competition tracks, i.e., Visual Grounding in Real-world Scenarios (VG-RS), Visual Question Answering with Spatial Awareness (VQA-SA), and Visual Reasoning in Creative Advertisement Videos (VR-Ads). Finally, 76 teams from the renowned academic and industrial institutions have registered and 40+ valid submissions (out of 1200+) have been included in our ranking lists. Our datasets, code sets (40+ baselines and 15+ participants' methods), and rankings are publicly available on the MARS2 workshop website and our GitHub organization page https://github.com/mars2workshop/, where our updates and announcements of upcoming events will be continuously provided.

A Survey of Scientific Large Language Models: From Data Foundations to Agent Frontiers

Scientific Large Language Models (Sci-LLMs) are transforming how knowledge is represented, integrated, and applied in scientific research, yet their progress is shaped by the complex nature of scientific data. This survey presents a comprehensive, data-centric synthesis that reframes the development of Sci-LLMs as a co-evolution between models and their underlying data substrate. We formulate a unified taxonomy of scientific data and a hierarchical model of scientific knowledge, emphasizing the multimodal, cross-scale, and domain-specific challenges that differentiate scientific corpora from general natural language processing datasets. We systematically review recent Sci-LLMs, from general-purpose foundations to specialized models across diverse scientific disciplines, alongside an extensive analysis of over 270 pre-/post-training datasets, showing why Sci-LLMs pose distinct demands -- heterogeneous, multi-scale, uncertainty-laden corpora that require representations preserving domain invariance and enabling cross-modal reasoning. On evaluation, we examine over 190 benchmark datasets and trace a shift from static exams toward process- and discovery-oriented assessments with advanced evaluation protocols. These data-centric analyses highlight persistent issues in scientific data development and discuss emerging solutions involving semi-automated annotation pipelines and expert validation. Finally, we outline a paradigm shift toward closed-loop systems where autonomous agents based on Sci-LLMs actively experiment, validate, and contribute to a living, evolving knowledge base. Collectively, this work provides a roadmap for building trustworthy, continually evolving artificial intelligence (AI) systems that function as a true partner in accelerating scientific discovery.

RetinaLogos: Fine-Grained Synthesis of High-Resolution Retinal Images Through Captions

The scarcity of high-quality, labelled retinal imaging data, which presents a significant challenge in the development of machine learning models for ophthalmology, hinders progress in the field. To synthesise Colour Fundus Photographs (CFPs), existing methods primarily relying on predefined disease labels face significant limitations. However, current methods remain limited, thus failing to generate images for broader categories with diverse and fine-grained anatomical structures. To overcome these challenges, we first introduce an innovative pipeline that creates a large-scale, synthetic Caption-CFP dataset comprising 1.4 million entries, called RetinaLogos-1400k. Specifically, RetinaLogos-1400k uses large language models (LLMs) to describe retinal conditions and key structures, such as optic disc configuration, vascular distribution, nerve fibre layers, and pathological features. Furthermore, based on this dataset, we employ a novel three-step training framework, called RetinaLogos, which enables fine-grained semantic control over retinal images and accurately captures different stages of disease progression, subtle anatomical variations, and specific lesion types. Extensive experiments demonstrate state-of-the-art performance across multiple datasets, with 62.07% of text-driven synthetic images indistinguishable from real ones by ophthalmologists. Moreover, the synthetic data improves accuracy by 10%-25% in diabetic retinopathy grading and glaucoma detection, thereby providing a scalable solution to augment ophthalmic datasets.

LECTOR: Summarizing E-book Reading Content for Personalized Student Support

Educational e-book platforms provide valuable information to teachers and researchers through two main sources: reading activity data and reading content data. While reading activity data is commonly used to analyze learning strategies and predict low-performing students, reading content data is often overlooked in these analyses. To address this gap, this study proposes LECTOR (Lecture slides and Topic Relationships), a model that summarizes information from reading content in a format that can be easily integrated with reading activity data. Our first experiment compared LECTOR to representative Natural Language Processing (NLP) models in extracting key information from 2,255 lecture slides, showing an average improvement of 5% in F1-score. These results were further validated through a human evaluation involving 28 students, which showed an average improvement of 21% in F1-score over a model predominantly used in current educational tools. Our second experiment compared reading preferences extracted by LECTOR with traditional reading activity data in predicting low-performing students using 600,712 logs from 218 students. The results showed a tendency to improve the predictive performance by integrating LECTOR. Finally, we proposed examples showing the potential application of the reading preferences extracted by LECTOR in designing personalized interventions for students.

Single-Agent vs. Multi-Agent LLM Strategies for Automated Student Reflection Assessment

We explore the use of Large Language Models (LLMs) for automated assessment of open-text student reflections and prediction of academic performance. Traditional methods for evaluating reflections are time-consuming and may not scale effectively in educational settings. In this work, we employ LLMs to transform student reflections into quantitative scores using two assessment strategies (single-agent and multi-agent) and two prompting techniques (zero-shot and few-shot). Our experiments, conducted on a dataset of 5,278 reflections from 377 students over three academic terms, demonstrate that the single-agent with few-shot strategy achieves the highest match rate with human evaluations. Furthermore, models utilizing LLM-assessed reflection scores outperform baselines in both at-risk student identification and grade prediction tasks. These findings suggest that LLMs can effectively automate reflection assessment, reduce educators' workload, and enable timely support for students who may need additional assistance. Our work emphasizes the potential of integrating advanced generative AI technologies into educational practices to enhance student engagement and academic success.

Attention Mamba: Time Series Modeling with Adaptive Pooling Acceleration and Receptive Field Enhancements

"This work has been submitted to the lEEE for possible publication. Copyright may be transferred without noticeafter which this version may no longer be accessible." Time series modeling serves as the cornerstone of real-world applications, such as weather forecasting and transportation management. Recently, Mamba has become a promising model that combines near-linear computational complexity with high prediction accuracy in time series modeling, while facing challenges such as insufficient modeling of nonlinear dependencies in attention and restricted receptive fields caused by convolutions. To overcome these limitations, this paper introduces an innovative framework, Attention Mamba, featuring a novel Adaptive Pooling block that accelerates attention computation and incorporates global information, effectively overcoming the constraints of limited receptive fields. Furthermore, Attention Mamba integrates a bidirectional Mamba block, efficiently capturing long-short features and transforming inputs into the Value representations for attention mechanisms. Extensive experiments conducted on diverse datasets underscore the effectiveness of Attention Mamba in extracting nonlinear dependencies and enhancing receptive fields, establishing superior performance among leading counterparts. Our codes will be available on GitHub.

GMAI-VL-R1: Harnessing Reinforcement Learning for Multimodal Medical Reasoning

Recent advances in general medical AI have made significant strides, but existing models often lack the reasoning capabilities needed for complex medical decision-making. This paper presents GMAI-VL-R1, a multimodal medical reasoning model enhanced by reinforcement learning (RL) to improve its reasoning abilities. Through iterative training, GMAI-VL-R1 optimizes decision-making, significantly boosting diagnostic accuracy and clinical support. We also develop a reasoning data synthesis method, generating step-by-step reasoning data via rejection sampling, which further enhances the model's generalization. Experimental results show that after RL training, GMAI-VL-R1 excels in tasks such as medical image diagnosis and visual question answering. While the model demonstrates basic memorization with supervised fine-tuning, RL is crucial for true generalization. Our work establishes new evaluation benchmarks and paves the way for future advancements in medical reasoning models. Code, data, and model will be released at \href{https://github.com/uni-medical/GMAI-VL-R1}{this link}.

An explainable transformer circuit for compositional generalization

Compositional generalization-the systematic combination of known components into novel structures-remains a core challenge in cognitive science and machine learning. Although transformer-based large language models can exhibit strong performance on certain compositional tasks, the underlying mechanisms driving these abilities remain opaque, calling into question their interpretability. In this work, we identify and mechanistically interpret the circuit responsible for compositional induction in a compact transformer. Using causal ablations, we validate the circuit and formalize its operation using a program-like description. We further demonstrate that this mechanistic understanding enables precise activation edits to steer the model's behavior predictably. Our findings advance the understanding of complex behaviors in transformers and highlight such insights can provide a direct pathway for model control.