@inproceedings{roy-etal-2026-prochunker,
title = "{P}ro{C}hunker: Self-Optimizing Procedural Chunking for Agent-Ready Workflow Graphs",
author = "Roy, Soumajit and
Adhikari, Sayanta and
Puranik, Vinayak S and
Comar, Prakash Mandayam",
editor = "Atanasova, Pepa and
Haffari, Gholamreza and
Wang, Minghan",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-industry.208/",
pages = "3490--3519",
ISBN = "979-8-89176-460-6",
abstract = "Can an LLM read a 30+ page Standard Operating Procedure (SOP) and produce a team of agents that reliably execute it? Industrial SOPs bury workflow logic in prose, leave inter-step dependencies implicit, and scatter referenced entities across sections, forcing planners to parse, sequence, and tool-map within one context window. We argue that the missing primitive is \textit{procedural chunking}: decomposing a document into self-contained executable units. Unlike RAG-style chunking, each chunk must be independently executable and chunks must jointly form a valid DAG encoding data and process dependencies, turning a prose SOP into a compact orchestrable workflow graph. We present \textit{ProChunker}, a three-phase self-optimizing LLM pipeline (Boundary Detection, Metadata Extraction, Dependency Construction) with \textit{Stochastic Best-of-N} prompt exploration guided by five custom metrics (Boundary Clarity, Entity Reference Coverage, Agent Self-Containedness Agreement, Step Coverage, Graph Validity); we then fine-tune Qwen3-1.7B on the pipeline{'}s own phase outputs, cutting latency by $\sim$27 min. Across three corpora (500 BioProBench protocols, six Industrial SOPs, 14 SOP-Bench documents), \textit{ProChunker} improves the five-metric composite over the strongest baseline by $+0.036$ to $+0.111$ across corpora. Our study shows ProChunker increases agent construction success rate by $+17\%$ and resolves ${>}90\%$ of procedural ambiguity."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="roy-etal-2026-prochunker">
<titleInfo>
<title>ProChunker: Self-Optimizing Procedural Chunking for Agent-Ready Workflow Graphs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Soumajit</namePart>
<namePart type="family">Roy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sayanta</namePart>
<namePart type="family">Adhikari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vinayak</namePart>
<namePart type="given">S</namePart>
<namePart type="family">Puranik</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Prakash</namePart>
<namePart type="given">Mandayam</namePart>
<namePart type="family">Comar</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: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pepa</namePart>
<namePart type="family">Atanasova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gholamreza</namePart>
<namePart type="family">Haffari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minghan</namePart>
<namePart type="family">Wang</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">979-8-89176-460-6</identifier>
</relatedItem>
<abstract>Can an LLM read a 30+ page Standard Operating Procedure (SOP) and produce a team of agents that reliably execute it? Industrial SOPs bury workflow logic in prose, leave inter-step dependencies implicit, and scatter referenced entities across sections, forcing planners to parse, sequence, and tool-map within one context window. We argue that the missing primitive is procedural chunking: decomposing a document into self-contained executable units. Unlike RAG-style chunking, each chunk must be independently executable and chunks must jointly form a valid DAG encoding data and process dependencies, turning a prose SOP into a compact orchestrable workflow graph. We present ProChunker, a three-phase self-optimizing LLM pipeline (Boundary Detection, Metadata Extraction, Dependency Construction) with Stochastic Best-of-N prompt exploration guided by five custom metrics (Boundary Clarity, Entity Reference Coverage, Agent Self-Containedness Agreement, Step Coverage, Graph Validity); we then fine-tune Qwen3-1.7B on the pipeline’s own phase outputs, cutting latency by \sim27 min. Across three corpora (500 BioProBench protocols, six Industrial SOPs, 14 SOP-Bench documents), ProChunker improves the five-metric composite over the strongest baseline by +0.036 to +0.111 across corpora. Our study shows ProChunker increases agent construction success rate by +17% and resolves >90% of procedural ambiguity.</abstract>
<identifier type="citekey">roy-etal-2026-prochunker</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-industry.208/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>3490</start>
<end>3519</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ProChunker: Self-Optimizing Procedural Chunking for Agent-Ready Workflow Graphs
%A Roy, Soumajit
%A Adhikari, Sayanta
%A Puranik, Vinayak S.
%A Comar, Prakash Mandayam
%Y Atanasova, Pepa
%Y Haffari, Gholamreza
%Y Wang, Minghan
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-460-6
%F roy-etal-2026-prochunker
%X Can an LLM read a 30+ page Standard Operating Procedure (SOP) and produce a team of agents that reliably execute it? Industrial SOPs bury workflow logic in prose, leave inter-step dependencies implicit, and scatter referenced entities across sections, forcing planners to parse, sequence, and tool-map within one context window. We argue that the missing primitive is procedural chunking: decomposing a document into self-contained executable units. Unlike RAG-style chunking, each chunk must be independently executable and chunks must jointly form a valid DAG encoding data and process dependencies, turning a prose SOP into a compact orchestrable workflow graph. We present ProChunker, a three-phase self-optimizing LLM pipeline (Boundary Detection, Metadata Extraction, Dependency Construction) with Stochastic Best-of-N prompt exploration guided by five custom metrics (Boundary Clarity, Entity Reference Coverage, Agent Self-Containedness Agreement, Step Coverage, Graph Validity); we then fine-tune Qwen3-1.7B on the pipeline’s own phase outputs, cutting latency by \sim27 min. Across three corpora (500 BioProBench protocols, six Industrial SOPs, 14 SOP-Bench documents), ProChunker improves the five-metric composite over the strongest baseline by +0.036 to +0.111 across corpora. Our study shows ProChunker increases agent construction success rate by +17% and resolves >90% of procedural ambiguity.
%U https://aclanthology.org/2026.emnlp-industry.208/
%P 3490-3519
Markdown (Informal)
[ProChunker: Self-Optimizing Procedural Chunking for Agent-Ready Workflow Graphs](https://aclanthology.org/2026.emnlp-industry.208/) (Roy et al., EMNLP 2026)
ACL