We describe our system for the ArchEHR-QA Shared Task on answering clinical questions using electronic health records (EHRs). Our approach uses large language models in two steps: first, to find sentences in the EHR relevant to a clinician's question, and second, to generate a short, citation-supported response based on those sentences. We use few-shot prompting, self-consistency, and thresholding to improve the sentence classification step to decide which sentences are essential. We compare several models and find that a smaller 8B model performs better than a larger 70B model for identifying relevant information. Our results show that accurate sentence selection is critical for generating high-quality responses and that self-consistency with thresholding helps make these decisions more reliable.

We present ReXVQA, the largest and most comprehensive benchmark for visual question answering (VQA) in chest radiology, comprising approximately 696,000 questions paired with 160,000 chest X-rays studies across training, validation, and test sets. Unlike prior efforts that rely heavily on template based queries, ReXVQA introduces a diverse and clinically authentic task suite reflecting five core radiological reasoning skills: presence assessment, location analysis, negation detection, differential diagnosis, and geometric reasoning. We evaluate eight state-of-the-art multimodal large language models, including MedGemma-4B-it, Qwen2.5-VL, Janus-Pro-7B, and Eagle2-9B. The best-performing model (MedGemma) achieves 83.24% overall accuracy. To bridge the gap between AI performance and clinical expertise, we conducted a comprehensive human reader study involving 3 radiology residents on 200 randomly sampled cases. Our evaluation demonstrates that MedGemma achieved superior performance (83.84% accuracy) compared to human readers (best radiology resident: 77.27%), representing a significant milestone where AI performance exceeds expert human evaluation on chest X-ray interpretation. The reader study reveals distinct performance patterns between AI models and human experts, with strong inter-reader agreement among radiologists while showing more variable agreement patterns between human readers and AI models. ReXVQA establishes a new standard for evaluating generalist radiological AI systems, offering public leaderboards, fine-grained evaluation splits, structured explanations, and category-level breakdowns. This benchmark lays the foundation for next-generation AI systems capable of mimicking expert-level clinical reasoning beyond narrow pathology classification. Our dataset will be open-sourced at https://huggingface.co/datasets/rajpurkarlab/ReXVQA

B. VLM Fine-Tuning and Hierarchical QA Inference For the fine-tuning and inference stages, a compact VLM is employed. All trainable parameters are fine-tuned on the collected dataset to ensure efficient adaptation. The fine-tuned weights are chosen based on the highest VQA accuracy achieved on the validation set. A hierarchical questioning technique is used to mitigate the limitations of smaller VLMs in generating long paragraphs. Instead of generating a single long paragraph, the system asks structured questions in a hierarchical manner, enabling efficient processing and resource minimization. This allows the VLM to generate meaningful scene descriptions without compromising speed or scalability. The hierarchical QA strategy optimizes inference time by dynamically selecting relevant questions based on the visual elements in the scene. For example, if the question "Is the ego vehicle moving on a straight road?" is answered affirmatively, subsequent questions such as "In which direction does the road curve?" are automatically answered as "none," reducing unnecessary computations. Similarly, if the answer to "Is it possible for the ego vehicle to turn right at the intersection?" is "no," other related questions, such as "Is there a vehicle approaching the intersection from the opposite direction?"

Multi-hop question generation (MQG) aims to generate questions that require synthesizing multiple information snippets from documents to derive target answers. The primary challenge lies in effectively pinpointing crucial information snippets related to question-answer (QA) pairs, typically relying on keywords. However, existing works fail to fully utilize the guiding potential of keywords and neglect to differentiate the distinct roles of question-specific and document-specific keywords. To address this, we define dual-perspective keywords (i.e., question and document keywords) and propose a Dual-Perspective Keyword-Guided (DPKG) framework, which seamlessly integrates keywords into the multi-hop question generation process. We argue that question keywords capture the questioner's intent, whereas document keywords reflect the content related to the QA pair. Functionally, question and document keywords work together to pinpoint essential information snippets in the document, with question keywords required to appear in the generated question. The DPKG framework consists of an expanded transformer encoder and two answer-aware transformer decoders for keyword and question generation, respectively. Extensive experiments demonstrate the effectiveness of our work, showcasing its promising performance and underscoring its significant value in the MQG task.

Querying generative AI models, e.g., large language models (LLMs), has become a prevalent method for information acquisition. However, existing query-answer datasets primarily focus on textual responses, making it challenging to address complex user queries that require visual demonstrations or explanations for better understanding. To bridge this gap, we construct a benchmark, RealVideoQuest, designed to evaluate the abilities of text-to-video (T2V) models in answering real-world, visually grounded queries. It identifies 7.5K real user queries with video response intents from Chatbot-Arena and builds 4.5K high-quality query-video pairs through a multistage video retrieval and refinement process. We further develop a multi-angle evaluation system to assess the quality of generated video answers. Experiments indicate that current T2V models struggle with effectively addressing real user queries, pointing to key challenges and future research opportunities in multimodal AI.

Structured scene representations are a core component of embodied agents, helping to consolidate raw sensory streams into readable, modular, and searchable formats. Due to their high computational overhead, many approaches build such representations in advance of the task. However, when the task specifications change, such static approaches become inadequate as they may miss key objects, spatial relations, and details. We introduce GraphPad, a modifiable structured memory that an agent can tailor to the needs of the task through API calls. It comprises a mutable scene graph representing the environment, a navigation log indexing frame-by-frame content, and a scratchpad for task-specific notes. Together, GraphPad serves as a dynamic workspace that remains complete, current, and aligned with the agent's immediate understanding of the scene and its task. On the OpenEQA benchmark, GraphPad attains 55.3%, a +3.0% increase over an image-only baseline using the same vision-language model, while operating with five times fewer input frames. These results show that allowing online, language-driven refinement of 3-D memory yields more informative representations without extra training or data collection.

Automatically generating high-quality mathematical problems that align with educational objectives is a crucial task in NLP-based educational technology. Traditional generation methods focus primarily on textual quality, but they often overlook educational objectives. Moreover, these methods address only single-dimensional, simple question generation, failing to meet complex, multifaceted educational requirements. To address these challenges, we constructed and annotated EduMath, a dataset of 16k mathematical questions with multi-dimensional educational objectives. Based on this dataset, we developed EQGEVAL, which incorporates three evaluation dimensions and is designed to assess the ability of models to generate educational questions. Drawing inspiration from teachers' problem design processes, we propose the Educational Question Planning with self-Reflection (EQPR) method for educational mathematical question generation, following a "plan-evaluate-optimize" approach. Specifically, by combining planning algorithm based on Monte Carlo Tree Search with the generative capabilities of Large Language Models, we continuously optimize questions through iterative feedback. This self-optimization mechanism ensures that the generated questions both fit the educational context and strategically achieve specific basic educational objectives. Through extensive experiments based on EQGEVAL, we have demonstrated that EQPR achieves significant improvements in generating questions that meet multi-dimensional educational objectives.

Multi-answer question answering (QA), where questions can have many valid answers, presents a significant challenge for existing retrieval-augmented generation-based QA systems, as these systems struggle to retrieve and then synthesize a large number of evidence passages. To tackle these challenges, we propose a new multi-answer QA framework -- Retrieval-augmented Independent Reading with Inter-passage Verification (RI$^2$VER). Our framework retrieves a large set of passages and processes each passage individually to generate an initial high-recall but noisy answer set. Then we propose a new inter-passage verification pipeline that validates every candidate answer through (1) Verification Question Generation, (2) Gathering Additional Evidence, and (3) Verification with inter-passage synthesis. Evaluations on the QAMPARI and RoMQA datasets demonstrate that our framework significantly outperforms existing baselines across various model sizes, achieving an average F1 score improvement of 11.17%. Further analysis validates that our inter-passage verification pipeline enables our framework to be particularly beneficial for questions requiring multi-evidence synthesis.

Retrieval-Augmented Generation (RAG) systems are increasingly diverse, yet many suffer from monolithic designs that tightly couple core functions like query reformulation, retrieval, reasoning, and verification. This limits their interpretability, systematic evaluation, and targeted improvement, especially for complex multi-hop question answering. We introduce ComposeRAG, a novel modular abstraction that decomposes RAG pipelines into atomic, composable modules. Each module, such as Question Decomposition, Query Rewriting, Retrieval Decision, and Answer Verification, acts as a parameterized transformation on structured inputs/outputs, allowing independent implementation, upgrade, and analysis. To enhance robustness against errors in multi-step reasoning, ComposeRAG incorporates a self-reflection mechanism that iteratively revisits and refines earlier steps upon verification failure. Evaluated on four challenging multi-hop QA benchmarks, ComposeRAG consistently outperforms strong baselines in both accuracy and grounding fidelity. Specifically, it achieves up to a 15% accuracy improvement over fine-tuning-based methods and up to a 5% gain over reasoning-specialized pipelines under identical retrieval conditions. Crucially, ComposeRAG significantly enhances grounding: its verification-first design reduces ungrounded answers by over 10% in low-quality retrieval settings, and by approximately 3% even with strong corpora. Comprehensive ablation studies validate the modular architecture, demonstrating distinct and additive contributions from each component. These findings underscore ComposeRAG's capacity to deliver flexible, transparent, scalable, and high-performing multi-hop reasoning with improved grounding and interpretability.

Existing RAG benchmarks often overlook query difficulty, leading to inflated performance on simpler questions and unreliable evaluations. A robust benchmark dataset must satisfy three key criteria: quality, diversity, and difficulty, which capturing the complexity of reasoning based on hops and the distribution of supporting evidence. In this paper, we propose MHTS (Multi-Hop Tree Structure), a novel dataset synthesis framework that systematically controls multi-hop reasoning complexity by leveraging a multi-hop tree structure to generate logically connected, multi-chunk queries. Our fine-grained difficulty estimation formula exhibits a strong correlation with the overall performance metrics of a RAG system, validating its effectiveness in assessing both retrieval and answer generation capabilities. By ensuring high-quality, diverse, and difficulty-controlled queries, our approach enhances RAG evaluation and benchmarking capabilities.