Specific learning disability in mathematics is the weakest persona profile across all GRPO-trained models¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? Breaking persona-aware Fit by disability type shows specific learning disability in mathematics is the weakest profile across every GRPO-trained model; even M7 drops below 7.8 on sld_math overall Fit. → Unggi Lee 2026

Evidence¶

Unggi Lee 2026¶

Unggi Lee, Jihoi Na, Yeil Jeong, Haeun Park, Yeonju Jang. (2026). Reinforcement Learning for Special Education: Aligning LLM Tutors to Diverse Learners through Disability-Adaptive Training. https://arxiv.org/abs/2605.30670 doi:10.48550/arxiv.2605.30670

q2 · i? · design · r2

Per-disability-type breakdown of persona-aware Fit scores on SPED-Tutor-Mix reported in the results section. The authors attribute the weakness to a comparatively small number of LD-appropriate mathematics items and a rubric emphasizing alternative representations a text-only tutor cannot deliver.

"Breaking the persona-aware Fit score by disability type reveals thatspecific learning disability in mathematics( sld_math) is the weakest profile across every GRPO-trained model. Even M7, which leads on every other profile, drops below 7.8 on sld_math overall Fit."

Discussion¶

Related Claims¶

• The full Special-R1 model raises persona-aware Fit from 6.75 to 8.40 and SPED-rubric Helpfulness from 0.720 to 0.768 over a non-disability-aware baseline — related
• Prompt style is the largest driver of disability-aware Fit gains, and the Thinking Reward becomes effective only in combination with adaptive prompting — related