@inproceedings{tian-etal-2026-think,
title = "Think Before You Score: Pedagogical Chain-of-Thought Annotation and Reinforcement Learning for Low-Resource Multilingual Essay Evaluation",
author = "Tian, Wei and
Zhou, Yuhao and
Li, Shenwei and
Sun, Xiayu and
Lan, Man",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.883/",
pages = "18641--18660",
ISBN = "979-8-89176-462-0",
abstract = "Automated Essay Scoring (AES) in low-resource, non-English educational settings faces a compounded challenge: existing corpora record only a final holistic score, discarding the rich intermediate reasoning that human examiners rely on. We argue this annotation gap may contribute to Rubric Drift{---}the observed tendency of LLMs to apply scoring norms misaligned with the target curriculum, potentially reflecting Anglocentric defaults. To address this, we introduce MPEAB-2430, a multilingual AES benchmark spanning Malay, Hindi, and Arabic curricula, annotated with Pedagogical Chain-of-Thought (P-CoT) trajectories that capture sub-dimensional analyses, explicit error attribution, and curriculum-conformant feedback, derived entirely from independent, language-specific expert practice. Leveraging this rich but modest-scale supervision, we propose RL-CRAF++, a sample-efficient pipeline combining CoT-SFT with Constrained GRPO. Crucially, our RL formulation relies entirely on human-annotation-derived rewards, completely bypassing the potential biases of external LLM judges. Evaluated independently per language, a 4B parameter RL-CRAF++ model consistently surpasses zero-shot frontier models (including GPT-5 and DeepSeek-V3.2) on in-domain test sets. Ablation analysis reveals that P-CoT supervision accounts for approximately 78.1{\%} of the alignment gain over standard SFT, providing strong evidence that, in this culturally situated evaluation setting, process-level structural supervision contributes substantially more than RL-stage calibration alone. Our code is available at https://github.com/TW-NLP/MPEAB"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tian-etal-2026-think">
<titleInfo>
<title>Think Before You Score: Pedagogical Chain-of-Thought Annotation and Reinforcement Learning for Low-Resource Multilingual Essay Evaluation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Tian</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuhao</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shenwei</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiayu</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Man</namePart>
<namePart type="family">Lan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Automated Essay Scoring (AES) in low-resource, non-English educational settings faces a compounded challenge: existing corpora record only a final holistic score, discarding the rich intermediate reasoning that human examiners rely on. We argue this annotation gap may contribute to Rubric Drift—the observed tendency of LLMs to apply scoring norms misaligned with the target curriculum, potentially reflecting Anglocentric defaults. To address this, we introduce MPEAB-2430, a multilingual AES benchmark spanning Malay, Hindi, and Arabic curricula, annotated with Pedagogical Chain-of-Thought (P-CoT) trajectories that capture sub-dimensional analyses, explicit error attribution, and curriculum-conformant feedback, derived entirely from independent, language-specific expert practice. Leveraging this rich but modest-scale supervision, we propose RL-CRAF++, a sample-efficient pipeline combining CoT-SFT with Constrained GRPO. Crucially, our RL formulation relies entirely on human-annotation-derived rewards, completely bypassing the potential biases of external LLM judges. Evaluated independently per language, a 4B parameter RL-CRAF++ model consistently surpasses zero-shot frontier models (including GPT-5 and DeepSeek-V3.2) on in-domain test sets. Ablation analysis reveals that P-CoT supervision accounts for approximately 78.1% of the alignment gain over standard SFT, providing strong evidence that, in this culturally situated evaluation setting, process-level structural supervision contributes substantially more than RL-stage calibration alone. Our code is available at https://github.com/TW-NLP/MPEAB</abstract>
<identifier type="citekey">tian-etal-2026-think</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.883/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>18641</start>
<end>18660</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Think Before You Score: Pedagogical Chain-of-Thought Annotation and Reinforcement Learning for Low-Resource Multilingual Essay Evaluation
%A Tian, Wei
%A Zhou, Yuhao
%A Li, Shenwei
%A Sun, Xiayu
%A Lan, Man
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F tian-etal-2026-think
%X Automated Essay Scoring (AES) in low-resource, non-English educational settings faces a compounded challenge: existing corpora record only a final holistic score, discarding the rich intermediate reasoning that human examiners rely on. We argue this annotation gap may contribute to Rubric Drift—the observed tendency of LLMs to apply scoring norms misaligned with the target curriculum, potentially reflecting Anglocentric defaults. To address this, we introduce MPEAB-2430, a multilingual AES benchmark spanning Malay, Hindi, and Arabic curricula, annotated with Pedagogical Chain-of-Thought (P-CoT) trajectories that capture sub-dimensional analyses, explicit error attribution, and curriculum-conformant feedback, derived entirely from independent, language-specific expert practice. Leveraging this rich but modest-scale supervision, we propose RL-CRAF++, a sample-efficient pipeline combining CoT-SFT with Constrained GRPO. Crucially, our RL formulation relies entirely on human-annotation-derived rewards, completely bypassing the potential biases of external LLM judges. Evaluated independently per language, a 4B parameter RL-CRAF++ model consistently surpasses zero-shot frontier models (including GPT-5 and DeepSeek-V3.2) on in-domain test sets. Ablation analysis reveals that P-CoT supervision accounts for approximately 78.1% of the alignment gain over standard SFT, providing strong evidence that, in this culturally situated evaluation setting, process-level structural supervision contributes substantially more than RL-stage calibration alone. Our code is available at https://github.com/TW-NLP/MPEAB
%U https://aclanthology.org/2026.findings-emnlp.883/
%P 18641-18660
Markdown (Informal)
[Think Before You Score: Pedagogical Chain-of-Thought Annotation and Reinforcement Learning for Low-Resource Multilingual Essay Evaluation](https://aclanthology.org/2026.findings-emnlp.883/) (Tian et al., Findings 2026)
ACL