@inproceedings{du-etal-2026-criticgen,
title = "{C}ritic{G}en: Generation-Aware Evaluation as Actionable Feedback",
author = "Du, Huifang and
Zuo, Zecheng and
Wang, Sen and
Wang, Haofen and
Yang, Yehui",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the {A}sia-Pacific Chapter of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.aacl-long.178/",
pages = "3975--3990",
ISBN = "979-8-89176-439-2",
abstract = "Current evaluation methods for large language models are coarse-grained and decoupled from generation, producing generic explanations that fail to provide actionable feedback for model improvement. We propose CriticGen, a fine-grained, generation-aware evaluation framework that turns evaluation into actionable control for answer improvement. CriticGen first generates sample-specific evaluation dimensions and scoring criteria, then uses the induced rubric to jointly produce a score, a reason, an executable refinement suggestion, and a refined answer. Experiments show that fine-grained evaluation should be both instance-specific and actionable. CriticGen improves human-rated rubric relevance/coverage from 3.33/4.03 to 3.97/4.24. On the score-consensus-filtered \textsc{RefineData} test set, the best CriticGen variant reaches 0.9556/0.9560 Pearson/Spearman agreement with the reference scores and improves reason/suggestion F1 to 0.7554/0.7900. Independent validation provides additional support: on 180 human-scored instances, CriticGen-Qwen2.5-7B obtains higher observed score correlations than GPT-5, with 0.8868/0.8969 Pearson/Spearman versus 0.8542/0.8615. Crucially, its feedback translates into substantial answer improvement, raising accuracy from 45.36{\%} to 91.26{\%} on 5,000 examples with reliable gold answers, with 47.30{\%} of answers corrected from Wrong$\rightarrow$Right and only 1.40{\%} degraded from Right$\rightarrow$Wrong."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="du-etal-2026-criticgen">
<titleInfo>
<title>CriticGen: Generation-Aware Evaluation as Actionable Feedback</title>
</titleInfo>
<name type="personal">
<namePart type="given">Huifang</namePart>
<namePart type="family">Du</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zecheng</namePart>
<namePart type="family">Zuo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sen</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haofen</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yehui</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-439-2</identifier>
</relatedItem>
<abstract>Current evaluation methods for large language models are coarse-grained and decoupled from generation, producing generic explanations that fail to provide actionable feedback for model improvement. We propose CriticGen, a fine-grained, generation-aware evaluation framework that turns evaluation into actionable control for answer improvement. CriticGen first generates sample-specific evaluation dimensions and scoring criteria, then uses the induced rubric to jointly produce a score, a reason, an executable refinement suggestion, and a refined answer. Experiments show that fine-grained evaluation should be both instance-specific and actionable. CriticGen improves human-rated rubric relevance/coverage from 3.33/4.03 to 3.97/4.24. On the score-consensus-filtered RefineData test set, the best CriticGen variant reaches 0.9556/0.9560 Pearson/Spearman agreement with the reference scores and improves reason/suggestion F1 to 0.7554/0.7900. Independent validation provides additional support: on 180 human-scored instances, CriticGen-Qwen2.5-7B obtains higher observed score correlations than GPT-5, with 0.8868/0.8969 Pearson/Spearman versus 0.8542/0.8615. Crucially, its feedback translates into substantial answer improvement, raising accuracy from 45.36% to 91.26% on 5,000 examples with reliable gold answers, with 47.30% of answers corrected from Wrong\rightarrowRight and only 1.40% degraded from Right\rightarrowWrong.</abstract>
<identifier type="citekey">du-etal-2026-criticgen</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-long.178/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>3975</start>
<end>3990</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CriticGen: Generation-Aware Evaluation as Actionable Feedback
%A Du, Huifang
%A Zuo, Zecheng
%A Wang, Sen
%A Wang, Haofen
%A Yang, Yehui
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-439-2
%F du-etal-2026-criticgen
%X Current evaluation methods for large language models are coarse-grained and decoupled from generation, producing generic explanations that fail to provide actionable feedback for model improvement. We propose CriticGen, a fine-grained, generation-aware evaluation framework that turns evaluation into actionable control for answer improvement. CriticGen first generates sample-specific evaluation dimensions and scoring criteria, then uses the induced rubric to jointly produce a score, a reason, an executable refinement suggestion, and a refined answer. Experiments show that fine-grained evaluation should be both instance-specific and actionable. CriticGen improves human-rated rubric relevance/coverage from 3.33/4.03 to 3.97/4.24. On the score-consensus-filtered RefineData test set, the best CriticGen variant reaches 0.9556/0.9560 Pearson/Spearman agreement with the reference scores and improves reason/suggestion F1 to 0.7554/0.7900. Independent validation provides additional support: on 180 human-scored instances, CriticGen-Qwen2.5-7B obtains higher observed score correlations than GPT-5, with 0.8868/0.8969 Pearson/Spearman versus 0.8542/0.8615. Crucially, its feedback translates into substantial answer improvement, raising accuracy from 45.36% to 91.26% on 5,000 examples with reliable gold answers, with 47.30% of answers corrected from Wrong\rightarrowRight and only 1.40% degraded from Right\rightarrowWrong.
%U https://aclanthology.org/2026.aacl-long.178/
%P 3975-3990
Markdown (Informal)
[CriticGen: Generation-Aware Evaluation as Actionable Feedback](https://aclanthology.org/2026.aacl-long.178/) (Du et al., AACL-IJCNLP 2026)
ACL
- Huifang Du, Zecheng Zuo, Sen Wang, Haofen Wang, and Yehui Yang. 2026. CriticGen: Generation-Aware Evaluation as Actionable Feedback. In Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3975–3990, Hengqin, China. Association for Computational Linguistics.