CoLearn: An Agentic Tutor that Learns its Learner in a Human--AI Co-Learning Loop

Kailai He; Zhihao Wu; Linhai Zhang; Runcong Zhao; Yulan He; Jiazheng Li

cs.CLSep 17, 2026

CoLearn: An Agentic Tutor that Learns its Learner in a Human--AI Co-Learning Loop

Authors: Kailai He, Zhihao Wu, Linhai Zhang, Runcong Zhao, Yulan He, Jiazheng Li

Organizations: King’s College London

Abstract

Good tutoring adapts to the individual: it tracks what a learner knows, notices why they go wrong, and asks the next question that will help most. Most deployed tutoring tools instead serve fixed item banks and treat a wrong answer as a single bit of signal. We present CoLearn, an interactive, agentic tutor that supports an iterative tutoring loop: the learner practises, and the system builds an evidence-grounded memory of the learner's mastery and misconceptions. This memory is updated as evidence accumulates and is used to generate the next personalised question. CoLearn has three components: (i) a persistent learner-state memory that updates per-topic mastery with a soft-evidence variant of Bayesian Knowledge Tracing, where a large language model acts as a continuous observation function; (ii) adaptive question generation that targets the learner's weakest topic and recurring misconceptions; and (iii) an evidence view that makes personalisation visible and testable through live progress visualisation and blind A/B comparison. In blind A/B evaluation, questions conditioned on this memory are preferred over non-personalised ones 68-69% of the time, and in persona simulations with hidden ground-truth mastery the agent's belief converges toward the learner's true mastery.

Figures & tables

Figure 1: CoLearn demo workflow across learner and administrator views. The learner first configures a practice session, answers an adaptive question, and receives diagnostic feedback with answer-linked annotations and live mastery estimates. The progress and administrator views expose the same personalization evidence through learner progress, blind A/B analytics, and agent convergence diagnostics.

Figure 2: The CoLearn co-learning loop. The learner answers; an LLM grades the answer into a continuous observation; soft-evidence BKT (Eq. 1 ) updates the agent’s cross-session memory $A_{t}$ ; the personalised generator reads that memory and a fixed strategy guideline to write the next, better-targeted question; and the loop repeats (the dashed self-evolve arrow is planned, not yet implemented). Two surfaces expose it: the learner view (live feedback and mastery) and the evidence view (progress dashboard + blind A/B).

Breakdown	$n$	Pref. rate	95% CI	$p$
Biology (overall)	18	0.68	[0.60, 0.77]	0.008
mixed learners	6	0.74	[0.71, 0.77]
strong learners	6	0.73	[0.59, 0.88]
Chemistry (overall)	18	0.69	[0.60, 0.77]	0.008
mixed learners	6	0.77	[0.61, 0.93]
strong learners	6	0.66	[0.52, 0.80]

Table 1: Blind A/B preference for personalised over non-personalised questions. Pref. rate is the mean per-rater preference, 95% CI is across raters, and

p

is a two-sided sign test vs.

0.5

. The mixed/strong rows are learner ability tiers (not subjects); weak-tier learners are excluded because, being weak on every skill, they cannot discriminate targeting, so the tier rows sum to fewer than the all row.

Metric	Adaptive	Random	Frozen
Weak-skill hit (warm)	0.72	0.55	0.57
Mixed learners	0.89	0.57	–
Strong learners	0.28	0.08	–
Mastery MAE $\downarrow$	0.12	0.16	0.20

Table 2: User-simulation (hidden ground-truth mastery). Weak-skill hit rows are from an 18-persona warm-start sweep; the Mastery MAE row is from the 6-persona Claude Sonnet 4.5-graded run. Weak-skill hit = fraction of served items on a truly-weak skill (higher = better targeting); Mastery MAE = final agent belief vs. true mastery (lower = better). Adaptive beats the Random-topic / Frozen controls on both. (Weak personas are weak on every skill, so they cannot discriminate targeting and are excluded from the per-tier rows.)

Figure 3: Memory convergence from the logged attempts ( $6$ personas, $132$ rounds, live framework grader Claude Sonnet 4.5). (a) Per-learner running mean belief error $|\hat{p}-p^{*}|$ under Adaptive : every learner’s estimate settles toward its hidden true mastery, most visibly the two strong learners that start far off. (b) Mean update magnitude $|\Delta p|$ by attempt (thin = per-attempt mean, bold = linear trend): under Adaptive it shrinks ( $0.14\rightarrow 0.10$ ) as the belief locks on, while the Random-topic and Frozen controls stay flat or drift upward.

Panel of Figure 4 — Figure 4: Per-item response distribution of the 1–5 Likert questionnaire over 16 raters. Blue = agree, grey = neutral, red = disagree. No item drew a strongly-disagree response; the weakest is the streamed-wait item (Q4), echoing the measured latency (§ 3 ).

Round type	Next/gen (s)	Answer (s)	Cost ($/rnd)
Short-answer	12.1	20.2	$\approx$ 0.015
Multiple-choice	7.6	0.01	$\approx$ 0.005

Table 3: Efficiency on the live backend (Bedrock Claude Sonnet 4.5). Short-answer

\approx

32 s/round (grade+feedback dominate); MCQ

\approx

7.6 s (deterministic answer). Cost is estimated from token usage.

Intelligent Tutoring Systems

Intelligent tutoring systems model learner knowledge and adapt instruction, practice, hints, and feedback to support learning.

See all 149 papers

DeepTutor: Towards Agentic Personalized Tutoring
Apr 10, 2026Bingxi Zhao, Jiahao Zhang, Xubin Ren +4LLM Personalization Intelligent Tutoring Systems
The University of Hong Kong
Hey Chat, Can You Teach Me? Structuring Socratic Dialogue for Human Learning in the Wild
Jun 10, 2026Sidney Tio, Arunesh Sinha, Pradeep VarakanthamIntelligent Tutoring Systems AI in Education
School of Computing and Information Systems Singapore Management University · Department of Management Science and Information Systems Rutgers Business School
RegulaTutor: A Theory-Informed Framework for Long-Term Personalized Learning
Oct 9, 2026Yonggang Li, Boqian Han, Rui Sheng +2Self-Regulated Learning Intelligent Tutoring Systems
The Hong Kong University of Science and Technology · The Chinese University of Hong Kong

Apr 10, 2026J/K move · Enter open · S save

DeepTutor: Towards Agentic Personalized Tutoring

Bingxi Zhao, Jiahao Zhang, Xubin Ren, Zirui Guo+3

The University of Hong Kong

LLM Personalization Intelligent Tutoring Systems User Modeling Memory-Augmented Language Model Agents

Open paper AskPDF arXiv

Education is one of the most promising real-world applications for Large Language Models (LLMs). However, current LLMs rely on static pre-training knowledge and lack adaptation to individual learners, while existing RAG systems fall short in delivering personalized, guided feedback. To bridge this gap, we present DeepTutor, a fully open-source agentic framework that unifies citation-grounded problem tutoring with difficulty-calibrated question generation. A hybrid personalization engine couples static knowledge grounding with dynamic learner memory, continuously adapting each interaction to the student's evolving needs. The same personalization substrate further extends to adaptive learning workflows, interactive books, and proactive multi-channel tutoring agents. To evaluate personalized tutoring, we introduce TutorBench, an interactive benchmark incorporating customized learner profiles grounded in university-level curricula across five domains. We further propose an LLM-based first-person interactive evaluation protocol that conducts assessments via a profile-driven student simulator. Complementary evaluations on established benchmarks, supported by human-alignment and ablation studies, confirm the framework's robustness and general utility. Results show that DeepTutor improves personalized metrics by 10.8% on average and strengthens general agentic reasoning across five backbone models by 29.4%.

Abstract

Intelligent Tutoring Systems

Explore similar work

DeepTutor: Towards Agentic Personalized Tutoring

Hey Chat, Can You Teach Me? Structuring Socratic Dialogue for Human Learning in the Wild

RegulaTutor: A Theory-Informed Framework for Long-Term Personalized Learning