# EduClaw-Bench: A Long-Horizon Benchmark for Pedagogical LLM Agents with Simulated Learners

arXiv: https://arxiv.org/abs/2608.03206 · HTML: https://mlsanity.com/paper/2608.03206/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Educational agents and math problem solving, one of 16 directions in LLM agents and tool use.
- Builds on 19 of its 35 references that are in the map.
- First posted on arXiv 4 Aug 2026.

## Placed in

- LLM agents and tool use / [Educational agents and math problem solving](https://mlsanity.com/agents/educational-agents-and-math-problem-solving/)

## Builds on (most cited first, 19 of 19)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [WebArena: A Realistic Web Environment for Building Autonomous Agents](https://mlsanity.com/paper/2307.13854/) (2023)
- [τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains](https://mlsanity.com/paper/2406.12045/) (2024)
- [OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments](https://mlsanity.com/paper/2404.07972/) (2024)
- [Context-Aware Attentive Knowledge Tracing](https://mlsanity.com/paper/2007.12324/) (2020)
- [MathDial: A Dialogue Tutoring Dataset with Rich Pedagogical Properties Grounded in Math Reasoning Problems](https://mlsanity.com/paper/2305.14536/) (2023)
- [CodeAid: Evaluating a Classroom Deployment of an LLM-based Programming Assistant that Balances Student and Educator Needs](https://mlsanity.com/paper/2401.11314/) (2024)
- [CodeHelp: Using Large Language Models with Guardrails for Scalable Support in Programming Classes](https://mlsanity.com/paper/2308.06921/) (2023)
- [Evaluating Language Models for Mathematics through Interactions](https://mlsanity.com/paper/2306.01694/) (2023)
- [Unifying AI Tutor Evaluation: An Evaluation Taxonomy for Pedagogical Ability Assessment of LLM-Powered AI Tutors](https://mlsanity.com/paper/2412.09416/) (2024)
- [Tutor CoPilot: A Human-AI Approach for Scaling Real-Time Expertise](https://mlsanity.com/paper/2410.03017/) (2024)
- [MathTutorBench: A Benchmark for Measuring Open-ended Pedagogical Capabilities of LLM Tutors](https://mlsanity.com/paper/2502.18940/) (2025)
- [Simulated Students in Tutoring Dialogues: Substance or Illusion?](https://mlsanity.com/paper/2601.04025/) (2026)
- [Instruct, Not Assist: LLM-based Multi-Turn Planning and Hierarchical Questioning for Socratic Code Debugging](https://mlsanity.com/paper/2406.11709/) (2024)
- [Automated Knowledge Concept Annotation and Question Representation Learning for Knowledge Tracing](https://mlsanity.com/paper/2410.01727/) (2024)
- [Evaluating Gemini in an arena for learning](https://mlsanity.com/paper/2505.24477/) (2025)
- [Evaluating Answer Leakage Robustness of LLM Tutors against Adversarial Student Attacks](https://mlsanity.com/paper/2604.18660/) (2026)
- [UniACE: A Unified Framework for Evaluating LLM Agentic Capabilities](https://mlsanity.com/paper/2605.27898/) (2026)
- [Rewarding How Models Think Pedagogically: Integrating Pedagogical Reasoning and Thinking Rewards for LLMs in Education](https://mlsanity.com/paper/2601.14560/) (2026)

## Built on by (most cited first, 1 of 1)

- [Do LLM Tutors Actually Teach? A Process-Centered Synthesis of the Tutoring Loop](https://mlsanity.com/paper/2610.15692/) (2026)

## Abstract (arXiv)

Large language models (LLMs) power educational applications from tutoring to essay scoring, but each is a point solution to a single task, and only recently have these point solutions been integrated into agents operating over a learning management system (LMS). Yet tutoring is long-horizon, since a learner improves over days and weeks rather than in a single turn, and no benchmark evaluates an agent tutor across a sustained relationship. We introduce EduClaw-Bench, a benchmark that places an agent tutor in a continuous 30-day relationship with a simulated learner grounded in knowledge tracing (KT), whose knowledge-concept mastery, from a KT model trained on real-student data, drives its answers and is probed for learning gain across 55 scenarios. Each agent is scored on three primary axes (learning gain, responsiveness, and helpfulness) and two curriculum-design axes (Gagn'e and Rosenshine), with helpfulness and the curriculum axes judged by a cross-family panel of three LLM judges. Evaluating 10 agent adapters over three base-model tiers yields two findings that single-tier, single-session evaluation cannot reach. First, tutoring quality belongs to the base model and the agent harness together rather than either alone. Second, almost no combination sustains good tutoring over the full horizon. A calibration check (ECE=0.049) and a live-classroom field study confirm that the simulated learner and its measurements track reality. Our work is a step toward trustworthy AI tutors for future education.