Knowledge-Based Visual Question Answering (KB-VQA) requires grounding visual queries to external knowledge beyond directly observable content in images.While recent multi modal large language models (MLLMs) show strong perceptual abilities, they struggle on KB-VQA tasks requiring groundings from both fine-grained entity and evidence levels.Most existing multi-modal retrieval augmented generation (MM-RAG) methods tightly couple entity discrimination and section-level evidence ranking into a single re-ranking stage, leading to high cost and limited generalization.In this work, we revisit existing MM-RAG solutions from a workflow perspective and argue both entity-level and fact-level groundings are key bottlenecks.We observe that although MLLMs often fail under open-ended entity naming, they can better identify the correct entity when selecting from a small set of candidate names.Based on this insight, we propose a simple and training-free identify-before-answer IBA framework that decouples entity identification from section-level re-ranking.Our approach prompts an MLLM to select high-confidence entities using only candidate names, followed by an off-the-shelf textual re-ranker for evidence selection.Experiments on Encyclopedic-VQA and InfoSeek show that our method consistently outperforms fine-tuned multi-modal re-ranking baselines while reducing training and inference complexity.Additional analyses reveal that the improvements arise not only from better entity identification, but also from selecting more informative evidence once correct entity is fixed.Our implementation is made public to ease reproducibility

In recent years, large language models (LLMs) have emerged as promising candidates for graph tasks. Many studies leverage natural language to describe graphs and apply LLMs for reasoning, yet most focus narrowly on performance benchmarks without fully comparing LLMs to graph learning models or exploring their broader potential. In this work, we present a comprehensive study of LLMs on graph learning tasks, evaluating both off-the-shelf and instruction-tuned models across a variety of scenarios. Beyond accuracy, we discuss data leakage concerns and computational overhead, and assess their performance under few-shot/zero-shot settings, domain transfer, structural understanding, and robustness. Our findings show that LLMs, particularly those with instruction tuning, greatly outperform traditional graph learning models in few-shot settings, exhibit strong domain transferability, and demonstrate excellent generalization and robustness. Our study highlights the broader capabilities of LLMs in graph learning and provides a foundation for future research.

Modern enterprises store vast, heterogeneous data across large data lakes, where the central challenge is shifting from answering an isolated question to discovering and assembling the tables and documents that together serve an analytical goal. We refer to such a coherent collection of related assets as a data product. Existing table-text benchmarks evaluate factoid question answering over individual tables and therefore do not capture this discovery and assembly problem. We introduce DPDisc, the first benchmark for data product discovery: given a high-level Data Product Request (DPR), a system must retrieve the coherent set of tables and passages that satisfy it. DPDisc is constructed by DPForge, an automated pipeline that repurposes table-text QA datasets by clustering related tables and passages into products, generating DPRs with an LLM ensemble, and validating them through multi-phase LLM evaluation. It comprises 13,076 validated instances with full provenance, spanning open-domain and financial sources. Sparse, dense, and hybrid baselines establish that the task is feasible yet reveal large performance gaps across domains, motivating structure-aware retrieval.

In-Context Learning (ICL) empowers Large Language Models (LLMs) with the ability to learn from a few examples provided in the prompt, enabling downstream generalization without the requirement for gradient updates. Despite encouragingly empirical success, the underlying mechanism of ICL remains unclear. Existing research remains ambiguous with various viewpoints, utilizing intuition-driven and ad-hoc technical solutions to interpret ICL. In this paper, we leverage a data generation perspective to reinterpret recent efforts from a systematic angle, demonstrating the potential broader usage of these popular technical solutions. For a conceptual definition, we rigorously adopt the terms of skill recognition and skill learning. Skill recognition selects one learned data generation function previously seen during pre-training while skill learning can learn new data generation functions from in-context data. Furthermore, we provide insights into the strengths and weaknesses of both abilities, emphasizing their commonalities through the perspective of data generation. This analysis suggests potential directions for future research. The corresponding paper list can be found here.