In math problem solving, a tutor-plus-peers LLM arrangement produced the strongest unassisted post-task test accuracy, with accuracy increasing monotonically from Control to Tutor + Peers¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r? · q3
Currency · newest study 2026 · 1 of 2 entries checked against their sources (automated) · not reviewed by a person

Subclaims¶

q3 i? Unassisted test accuracy increased monotonically across support conditions, with Tutor + Peers (~65%) highest and Control (~42%) lowest, and Tutor + Peers produced the largest significant gain over Control. → Harsh Kumar 2026

Evidence¶

Harsh Kumar 2026¶

Harsh Kumar, Zi Kang (Jace) Mu, Jonathan Vincentius, Ashton Anderson. (2026). Human Thinking under Plural LLM Assistance: Mathematical Problem Solving and Open-Ended Writing. Working Paper, July 2026, Toronto, Canada. https://arxiv.org/abs/2604.02677 doi:10.48550/arxiv.2604.02677

q3 · i? · causal · r?

Pre-registered controlled Prolific experiment (315 participants passing preregistered checks) with four between-subjects support conditions; test accuracy rose across conditions, with means "(∼42%), Peers Only (∼48%), Tutor Only (∼59%), and Tutor + Peers (∼65%)".

"Points show mean test accuracy (proportion correct in test round) across participants in each condition; vertical error bars indicate±1 SEM. (∼42%), Peers Only (∼48%), Tutor Only (∼59%), and Tutor + Peers (∼65%)."

Harsh Kumar 2026 (2)¶

Harsh Kumar, Zi Kang (Jace) Mu, Jonathan Vincentius, Ashton Anderson. (2026). Human Thinking under Plural LLM Assistance: Mathematical Problem Solving and Open-Ended Writing. Working Paper, July 2026, Toronto, Canada. https://arxiv.org/abs/2604.02677

q3 · i? · causal · r?

Same Experiment-1 test-phase analysis of isomorphic problem variants solved unassisted after a 1-minute distractor; the article reports "The Tutor + Peers condition produced the largest gain over Control" with the printed chi-square statistic and p-value.

"The Tutor + Peers condition produced the largest gain over Control (𝜒 2(1)= 14.83,𝑝<. 001)."

Discussion¶

Related Claims¶

• Adding a second writing agent increased reported cognitive load, and writing confidence did not increase monotonically with support, with the Single condition reporting the lowest confidence — reports the opposite
• In math problem solving, a tutor LLM alone significantly improved unassisted test accuracy over no-agent control, while peer-like error agents alone did not reach significance — related
• In a 10-student pilot, EduGuard produced higher post-test accuracy, larger normalized learning gain, lower calibration error, and lower overreliance than GPT-4o-mini Tutor — related
• AI access raises immediate unaided test scores by 7.2 percentage points (0.28 SD) after a timed learning phase — a broader claim this one bears on
• Unguarded answer-giving AI harmed unaided exam performance while a guarded version of the same model erased the harm (Bastani et al., 2025, as reported) — related
• In a randomized experiment on fraction problems, participants who used an AI assistant solved fewer unassisted test problems and skipped more than control participants — related
• Multi-turn teaching by LLM teacher agents yields larger student gains than knowledge-point exposure alone — related
• AI assistance impairs unassisted performance and persistence in reading comprehension, replicating the fraction-task effects in a second domain — related
• Students show large immediate pre-to-post knowledge gains within a single LLM-supported learning session — related
• Access to a structured conversational AI tutor increased post-test content scores over a no-AI holdout with ambient consumer AI available — related