The multi-step grading recipe with calibration examples achieves the highest scoring accuracy and receives the majority of bandit arm pulls¶
Claim · All claims
Evidence · 1 study · causalr1·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¶
Reading the evidence codes
q is the design tier, i the impact, n the sample, and the kind with its r
the rigour judged within that kind: what each code means.
This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.
q2 i? Multi-step assessment with calibration examples achieved the lowest MAE (approximately 0.85) and highest QWK (approximately 0.55), and the MAB allocated over 70% of pulls to it. → 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
Numerical results from the MAB experiment on 787 IELTS Task 2 essays scored with Gemini 2.5. The text reports "Multi-step Ex achieved the lowest error rate (approximately 0.85)"; QWK analysis showed approximately 0.55, and the bandit allocated approximately 350 pulls (over 70%) to this recipe.
"Figure 3 displays the Mean Absolute Error (MAE) for each recipe, revealing that Multi-step Ex achieved the lowest error rate (approximately 0.85), followed by Single-step Ex (1.0)."
Discussion¶
Related Claims¶
- The four grading recipes exhibit an accuracy-cost tradeoff, with the most accurate recipe costing roughly five times the balanced alternative — 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
- 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
- Bayesian adaptive design benefits study subjects by allocating them to more effective treatment arms — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
multi-step-examples-highest-aes-accuracy |
title |
The multi-step grading recipe with calibration examples achieves the highest scoring accuracy and receives the majority of bandit arm pulls |
description |
The multi-step grading recipe with calibration examples achieves the highest scoring accuracy and receives the majority of bandit arm pulls |
status |
review |
evidence_strength |
moderate |
newest_evidence |
2026 |
recheck_by |
2029 |
generated |
by process:wiki-ingest on 2026-10-10 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — olga-manakina-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.