Tag
humans& releases &Tutor, an AI model trained to teach by training against Persimmon as a simulated learner, achieving state-of-the-art results on the TutorMoments benchmark and outperforming frontier models like Claude Sonnet 5.5 and GPT-6 Astra under plain prompting.
Shares teaching practices from Wharton's "Generative AI Studio": treating generative AI as a creative medium rather than an efficiency tool, and using two classic art and design education methods — Crit (work critique) and Charrette (time-limited creation sprint) — to cultivate judgment, taste, creativity, and critical thinking, which are the most valuable traits in the AI era.
In his JPBM Communications Prize lecture, 3Blue1Brown founder Grant Sanderson uses the metaphor of King Midas's touch to argue that mathematics's precision is both a gift and a "pedagogical curse" — it makes rigor easy to measure, yet makes teaching clarity hard to verify. He therefore advocates "raising the ceiling of mathematical exposition while lowering the floor."
OmniEdu introduces an open family of foundation models for K-12 education, trained on a curated corpus to improve problem-solving, curriculum grounding, and pedagogical tutoring capabilities.
The paper proposes a five-layer architecture for AI-driven English textbooks, which was tested on 186 undergraduates, showing improved completion accuracy, speaking scores, and reduced teacher correction time.
This paper leverages offline reinforcement learning to automatically generate high-quality chess puzzles, enhancing pedagogical value for beginners based on extensive user data from platforms like Chess.com and Lichess.
TeachMateGPT presents a multi-agent framework that improves retrieval-augmented generation for creating pedagogical assessments from science textbooks, achieving higher faithfulness and answer relevancy compared to baseline systems.
This paper introduces the Pedagogical Suitability Index (PSI), a composite metric for evaluating and improving how well LLM-based AI tutors align responses with learner readiness and curricular progression, showing that PSI-guided feedback improves weak tutoring cases.
This design-based study investigates how pedagogical scaffolding can help ethnic minority preparatory students shift from passive consumption to critical co-creation with Generative AI in prompt engineering tasks, resulting in improved prompt self-efficacy and active gatekeeping of AI-generated content.
This paper presents a pre-registered audit of whether LLM-judged helpfulness can reliably distinguish answer-giving from pedagogical guidance in AI tutors. The authors find that general-purpose helpfulness is not a dependable pedagogy signal, recommending pedagogy-targeted rubrics and deterministic process measures instead.
An article from Wharton's Generative AI Studio describing how traditional arts pedagogy—studio structure, critique, and charrette—can be applied to teaching generative AI as a creative medium for developing products and services.
This paper presents HeuristicEdu, a pipeline to align Qwen2.5-7B as a Socratic tutor using supervised warm-up and GRPO with heuristic rewards, evaluated on a new dataset SocraticEdu, showing improved scaffolding effectiveness and reduced keyword leakage.
This paper characterizes the inherent interpretability of linear models vs. single-qubit mixed-state models for binary classification, showing that the quantum model learns a hyperellipsoid instead of a hyperplane, with implications for inductive biases and pedagogy.
LessonBench-V1 is a new benchmark dataset pairing 647 human-written STEM lessons with reverse-engineered lesson plans, enabling systematic evaluation of AI lesson generation agents.
Ello shares their engineering approach to building a real-time AI tutor for children ages 4-9, focusing on sub-second latency and pedagogy to avoid losing a child's attention.
Summary of three key takeaways from conversations with leading AI researchers at CAISconf, covering the importance of evaluations for AI agents, the trade-offs between industry and academia, and a novel pedagogical RL approach.
This paper discusses how to make sense of proof by contradiction, a fundamental mathematical reasoning technique, for educational purposes.
Introduces EduAgentBench, a source-grounded benchmark for evaluating tutor agents across professional pedagogical judgment, multi-turn tutoring, and autonomous teaching workflow execution. Evaluations on frontier models show they still fall short of professional teaching standards in situated tutoring and workflow tasks.
Describes a training technique involving spike-aware pedagogy rewards that penalize implausible jumps, and surprisal-gated imitation where the student learns easy tokens quickly and hard ones slowly.
The author discusses the lack of detailed, complete proofs in advanced mathematics textbooks, which creates unnecessary barriers for students and professionals, and advocates for the creation of more accessible accompaniment notes.