DeepTutor improves overall first-person interactive tutoring quality by 10.76% over a Naive Tutor baseline on TutorBench¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · checked against its sources 2026-10-20 (automated, 1 of 1 entry) · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? On TutorBench's first-person interactive evaluation, DeepTutor improves overall quality by 10.76% relative to the Naive Tutor baseline and leads on nearly all of ten metrics. → Bingxi Zhao 2026

Evidence¶

Bingxi Zhao 2026¶

Bingxi Zhao, Jiahao Zhang, Xubin Ren, Zirui Guo, Tianzhe Chu, Yi Ma, Chao Huang. (2026). DeepTutor: Towards Agentic Personalized Tutoring. Technical Report, The University of Hong Kong. https://arxiv.org/abs/2604.26962 doi:10.48550/arxiv.2604.26962

q2 · i? · design · r2

LLM-simulated interactive evaluation of DeepTutor versus four baselines (Naive, CoT, Self-Refine, ReAct) sharing the same backbone and RAG, over the full TutorBench release of 270 tasks. The article reports DeepTutor "improves overall quality by 10.76% and leads on nearly all metrics"; relative gain is not a standardized effect size.

"The four baselines remain tightly clustered, indicating that adding CoT, self-refinement, or ReAct-style tool use to the same backbone and RAG interface is not sufficient to matchDeepTutor's learner-adaptive behavior.DeepTutor improves overall quality by 10.76% and leads on nearly all metrics."

Discussion¶

Related Claims¶

• DeepTutor's interactive tutoring gains are stable across five university-level domains, varying only 0.16 points in overall quality — related
• Human raters' preferences align with the LLM judge on TutorBench (Pearson r = 0.82 across metric-level win rates) — related
• Tutoring quality belongs to the base model and agent harness together rather than either alone, as adapter rankings reorder across tiers — related
• TACTutor shows no degradation relative to its Qwen backbone on external educational benchmarks, with scores rising on MRBench, TutorBench, LongTutor, and DeepTutor — related
• Teacher-agent comparisons show backbone-dependent patterns consistent with ZPD-informed adaptation — related
• Education-oriented instruction tuning consistently improves curriculum grounding, K-12 problem solving, and pedagogical tutoring at every evaluated model scale — related
• Automated metrics and LLM judges can misalign with expert human preferences in non-verifiable educational evaluation — related
• Static Knowledge Grounding and Dynamic Personal Memory are complementary: removing both yields the largest quality degradation — related
• In a blinded study with 50 learners, TACTutor receives the highest overall mean rating (5.54/7) and is rated above its backbone across all four teaching-behavior dimensions (+0.39 overall) — related