Adaptation gains grow monotonically with model capacity under the same LoRA recipe, though three data points cannot establish a scaling law¶

Claim · All claims
Evidence · 1 study · causal 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? Accuracy improvements increase monotonically with parameter count (+5.5, +15.0, +23.3 pp from 0.6B to 4B), observed across accuracy, Token F1 and BERTScore. → Rashid Azraf Jahin 2026

Evidence¶

Rashid Azraf Jahin 2026¶

Rashid Azraf Jahin, Saadman Sajid, Khan Raiyan Ibne Reza, Sumaiya Tabassum Nimi. (2026). PhysicsMate: A Curriculum-Grounded Bengali Benchmark for Secondary Physics QA with Small-Model Adaptation. IEEE CSDE 2026. https://arxiv.org/abs/2610.00664 doi:10.48550/arxiv.2610.00664

q2 · i? · causal · r2

Numerical results reported with Fig. 2 across three Qwen3 scales under one fixed LoRA configuration (r = 16, α = 32, dropout 0.05). The authors state "Three data points is not enough to prove a scaling law" but the trend is monotonic across all three metrics.

"Figure 2 shows that adaptation gains grow with model capacity. The accuracy improvement trajectory is: ∆Acc : +5 .5 pp → +15.0 pp → +23.3 pp"

Discussion¶

Related Claims¶

• LoRA adaptation significantly increases closed-book accuracy on the PhysicsMate test set at all three Qwen3 scales — a narrower finding that bears on this claim
• CDPK performance scales with model size, with a sharp Pareto-frontier drop-off below around 8B parameters — a broader claim this one bears on
• Model family and instruction-tuning approach appear better predictors of tutoring quality than parameter count alone — related
• LoRA fine-tuning on control systems course data increases lexical and semantic similarity between model responses and course reference answers at both 3B and 7B model sizes — related
• Adaptation benefits structured curricular knowledge (quantities, laws, definitions) most and loosely specified entity-level knowledge least — related
• Under identical LoRA hyperparameters and training data, the 27B Gemma model outperforms the 70B LLaMA model on every computed essay-scoring metric — related
• Base-model size effects persist after LoRA adaptation: the 7B model gained more on ROUGE-2 and ROUGE-L, yet 3B-r16 outperformed 7B-r4 on ROUGE-L and BERTScore-F1 — related
• Scaling model size improves performance within K-5 and at the boundary but does not recover Beyond-K-5 performance — related