@inproceedings{rawal-etal-2026-benchmarking,
title = "Benchmarking Correctness and Security in Multi-Turn Code Generation",
author = "Rawal, Ruchit and
Chiang, Jeffrey Yang Fan and
Shen, Chihao and
Tian, Jeffery Siyuan and
Mahajan, Aastha and
Goldstein, Tom and
Chen, Yizheng",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1988/",
pages = "45222--45258",
ISBN = "979-8-89176-459-0",
abstract = "AI coding assistants powered by large language models (LLMs) have transformed software development, significantly boosting productivity. While existing benchmarks evaluate the correctness and security of LLM-generated code, they are limited to single-turn tasks that cannot capture the conversational nature of real-world development, where requirements evolve across turns and tools increasingly generate incremental diffs rather than complete programs. We introduce MT-Sec, the first benchmark to jointly evaluate correctness and security in multi-turn coding scenarios. MT-Sec covers four interaction types (expansion, editing, refactoring, and debugging) reflecting common development workflows. We construct multi-turn tasks using human-curated instructions and feedback-driven interactions grounded in dynamic test suites. Evaluating 31 LLMs and 3 agent scaffolds, we find that models struggle most with expansion and editing, but improve significantly in debugging when provided with feedback. Agent-based systems often underperform their base models in automated multi-turn settings due to brittle tool use. We also evaluate multi-turn code-diff generation, an unexplored but practical setting, and observe higher failure rates compared to full-program generation. Our findings underscore the need for multi-turn benchmarks to drive safer AI-assisted development."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="rawal-etal-2026-benchmarking">
<titleInfo>
<title>Benchmarking Correctness and Security in Multi-Turn Code Generation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ruchit</namePart>
<namePart type="family">Rawal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeffrey</namePart>
<namePart type="given">Yang</namePart>
<namePart type="given">Fan</namePart>
<namePart type="family">Chiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chihao</namePart>
<namePart type="family">Shen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeffery</namePart>
<namePart type="given">Siyuan</namePart>
<namePart type="family">Tian</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aastha</namePart>
<namePart type="family">Mahajan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tom</namePart>
<namePart type="family">Goldstein</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yizheng</namePart>
<namePart type="family">Chen</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>AI coding assistants powered by large language models (LLMs) have transformed software development, significantly boosting productivity. While existing benchmarks evaluate the correctness and security of LLM-generated code, they are limited to single-turn tasks that cannot capture the conversational nature of real-world development, where requirements evolve across turns and tools increasingly generate incremental diffs rather than complete programs. We introduce MT-Sec, the first benchmark to jointly evaluate correctness and security in multi-turn coding scenarios. MT-Sec covers four interaction types (expansion, editing, refactoring, and debugging) reflecting common development workflows. We construct multi-turn tasks using human-curated instructions and feedback-driven interactions grounded in dynamic test suites. Evaluating 31 LLMs and 3 agent scaffolds, we find that models struggle most with expansion and editing, but improve significantly in debugging when provided with feedback. Agent-based systems often underperform their base models in automated multi-turn settings due to brittle tool use. We also evaluate multi-turn code-diff generation, an unexplored but practical setting, and observe higher failure rates compared to full-program generation. Our findings underscore the need for multi-turn benchmarks to drive safer AI-assisted development.</abstract>
<identifier type="citekey">rawal-etal-2026-benchmarking</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1988/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>45222</start>
<end>45258</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Benchmarking Correctness and Security in Multi-Turn Code Generation
%A Rawal, Ruchit
%A Chiang, Jeffrey Yang Fan
%A Shen, Chihao
%A Tian, Jeffery Siyuan
%A Mahajan, Aastha
%A Goldstein, Tom
%A Chen, Yizheng
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F rawal-etal-2026-benchmarking
%X AI coding assistants powered by large language models (LLMs) have transformed software development, significantly boosting productivity. While existing benchmarks evaluate the correctness and security of LLM-generated code, they are limited to single-turn tasks that cannot capture the conversational nature of real-world development, where requirements evolve across turns and tools increasingly generate incremental diffs rather than complete programs. We introduce MT-Sec, the first benchmark to jointly evaluate correctness and security in multi-turn coding scenarios. MT-Sec covers four interaction types (expansion, editing, refactoring, and debugging) reflecting common development workflows. We construct multi-turn tasks using human-curated instructions and feedback-driven interactions grounded in dynamic test suites. Evaluating 31 LLMs and 3 agent scaffolds, we find that models struggle most with expansion and editing, but improve significantly in debugging when provided with feedback. Agent-based systems often underperform their base models in automated multi-turn settings due to brittle tool use. We also evaluate multi-turn code-diff generation, an unexplored but practical setting, and observe higher failure rates compared to full-program generation. Our findings underscore the need for multi-turn benchmarks to drive safer AI-assisted development.
%U https://aclanthology.org/2026.emnlp-main.1988/
%P 45222-45258
Markdown (Informal)
[Benchmarking Correctness and Security in Multi-Turn Code Generation](https://aclanthology.org/2026.emnlp-main.1988/) (Rawal et al., EMNLP 2026)
ACL
- Ruchit Rawal, Jeffrey Yang Fan Chiang, Chihao Shen, Jeffery Siyuan Tian, Aastha Mahajan, Tom Goldstein, and Yizheng Chen. 2026. Benchmarking Correctness and Security in Multi-Turn Code Generation. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 45222–45258, Budapest, Hungary. Association for Computational Linguistics.