The four grading recipes exhibit an accuracy-cost tradeoff, with the most accurate recipe costing roughly five times the balanced alternative¶

Claim · All claims
Evidence · 1 study · causal r1 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Multi-step Ex was most accurate (MAE 0.85) at about $0.0011 per essay, while Single-step Ex offered moderate accuracy (MAE 1.0) at $0.0003 and Single-step NoEx was most economical at $0.0002 with MAE 1.45. → Olga Manakina 2026

Evidence¶

Olga Manakina 2026¶

Olga Manakina, Igor Bogdanov. (2026). Learning to Grade Efficiently: A Bandit-Driven Prompt-Selection Framework for Low-Cost LLM Essay Scoring. https://arxiv.org/abs/2608.23814 doi:10.48550/arxiv.2608.23814

q2 · i? · causal · r1

Cost analysis plotting MAE against average estimated API cost per grading attempt for the four recipes (Figure 5). The article reports the accuracy-cost tradeoff, with Single-step NoEx the most economical at $0.0002 (MAE 1.45) and Multi-step NoEx worst on both metrics (MAE 1.55 at $0.0004).

"Multi-step Ex, while most accurate (MAE of 0.85), incurred a relatively high cost of approximately $0.0011 per essay. Single-step Ex offered a balanced alternative with moderate accuracy (MAE of 1.0) at a lower cost ($0.0003)."

Discussion¶

Related Claims¶

• The multi-step grading recipe with calibration examples achieves the highest scoring accuracy and receives the majority of bandit arm pulls — related
• Grading recipes without calibration examples perform substantially worse on MAE, though multi-step without examples ranks second on QWK — related
• Removing detailed rubric explanations from prompts improved grading accuracy across all approaches in an ablation study — related
• A multi-armed bandit controller achieves scoring accuracy comparable to exhaustive grid search while reducing LLM calls by 78.4% and token consumption by 72.8% — related