Tag
Ask-E is a new benchmark and training environment that evaluates and trains models on generating questions calibrated to specific skill levels, defined by the capabilities of two existing language models. Frontier models score below 50% on calibration, and training on Ask-E improves downstream math benchmarks without new math data or correctness-based rewards.
This paper evaluates six LLMs through Bloom's Taxonomy to assess their ability to generate educational questions that stimulate higher-order thinking, introducing a prompting strategy that reduces repetitiveness by 24.45% and increases higher-order outputs by 11.53%.
VeriGeo introduces a controllable geometry question generation framework that uses verification-guided reflection to ensure numerical and analytical consistency. The method produces high-quality synthetic data, achieving state-of-the-art results on GeoQA and strong performance on PGPS9K and MathVista-GPS.
This paper investigates how TMK-based question generation strategies affect dataset quality for procedural and multi-hop reasoning in AI learning systems, comparing strict TMK generation, transcript-first generation, and TMK-aware generation, and introduces a grounding validation framework.
This paper introduces HieraRAG, a hierarchical framework for determining optimal granularity in RAG benchmarks. It generates 5,872 synthetic QA pairs across three dimensions and finds that ideal granularity varies by dimension, offering a portable procedure for practitioners.
This paper introduces a self-evolving framework for vision-language models to improve their question-generation capabilities without external supervision, enhancing both question quality and answerer performance.
This paper presents the FETCH classifier, which uses an ensemble of LLMs to generate follow-up questions for automated legal intake, evaluating question quality and cost trade-offs. It finds that high-cost models like GPT-5 are needed for effective plain-language questions, and proposes a rubric for evaluating such questions.
This paper introduces slidesqaqa, a Flask-based software system that generates pedagogically useful questions from PDF slide decks. It uses a four-stage LLM pipeline to extract text and images, plan questions across the deck, annotate slides, and reconcile outputs, demonstrating high-fidelity question generation on technical lecture slides.