Do Recipes Have Personas? Characterizing and Generating Creator Style in Attributed Procedural Graphs

Lei Jiang

cs.AIAug 25, 2026

Do Recipes Have Personas? Characterizing and Generating Creator Style in Attributed Procedural Graphs

Authors: Lei Jiang

Organizations: Microsoft, Redmond WA 98052, USA

Abstract

While large language models (LLMs) possess vast zero-shot procedural knowledge, their tendency to produce homogenized logic often obscures the unique, idiosyncratic execution processes of individual human creators. In this paper, we investigate the computational discovery of procedural personas from unstructured data. To achieve this, we introduce ViralRecipesTrans, a new dataset of procedurally aligned execution flow graphs extracted from popular culinary video transcripts and explicitly mapped to specific creators. We formulate procedural stylometry as a graph learning and process discovery task, revealing a fundamental duality: while traditional lexical classifiers overfit via semantic leakage, discrete topological metrics successfully capture the rigid physical constraints of a creator's workflow. Building upon this characterization, we extend our framework into a novel generative task--predicting a creator's exact structural execution graph for unseen dishes. We expose a fundamental dichotomy in style generation between global macro-planning and local structural execution. Our results demonstrate that few-shot LLMs dominate semantic assignment but suffer from persistent macro-planning deficits, whereas our structured two-stage model achieves superior topological control via rigid Markovian priors. Together, an ensemble approach to procedural generation combines the strengths from both sides, dynamically synthesizing global semantic reasoning with localized topological footprints to automate the discovery and generation of personalized workflows.

Figures & tables

Metric	Long-form	Short-form	Overall
Cooking steps (mean $\pm$ std)	14.8 $\pm$ 10.6	9.3 $\pm$ 4.8	12.2 $\pm$ 8.9
Unique Ingredients (mean $\pm$ std)	13.2 $\pm$ 8.0	10.0 $\pm$ 5.2	11.7 $\pm$ 7.0
Transcript length (words, mean)	1800	176	1028
Transcript length (words, median)	1310	153	299
Global action vocabulary	548	499	675
Action vocab per account (mean)	106	89	97

Table 1: VRT per-recipe textual and topological statistics

Archetype Class ( $\mathcal{A}$ )	Precision	Recall	F1-Score
Written Text ( $a_{\text{text}}$ )	0.999	0.999	0.999
Viral Long-Form ( $a_{\text{long}}$ )	0.702	0.613	0.654
Viral Short-Form ( $a_{\text{short}}$ )	0.656	0.740	0.696
Overall Metric	Score
Accuracy	0.784
Macro F1	0.783

Table 2: Classification performance for the 3-class procedural archetype task. The perfect isolation of the written archetype confirms that cinematographic workflows possess a fundamentally distinct structural topology.

Figure 1: Scalability results as the number of classes increases

Figure 2: The workflow of the persona-controlled recipe generation task

	long videos			short videos
Item	IF1	nEF1	StepErr	IF1	nEF1	StepErr
Zero-shot LLM	0.223	0.338	0.481	0.261	0.341	0.458
Few-shot LLM	0.344	0.474	0.373	0.385	0.478	0.351
Autoregressive	0.318	0.473	0.446	0.353	0.446	0.410
N-gram sequence	0.303	0.452	0.549	0.363	0.446	0.513
Two-stage	0.294	0.456	0.350	0.348	0.455	0.332

Table 3: Performance comparison over the stylish recipe generation task

	Long Videos			Short Videos
Model	IF1	nEF1	StepErr $\downarrow$	IF1	nEF1	StepErr $\downarrow$
GPT-5.4	0.396	0.556	0.384	0.413	0.532	0.317
GPT-4.1-mini	0.393	0.504	0.376	0.481	0.446	0.269
Qwen3.6-27B	0.369	0.510	0.409	0.410	0.476	0.248

Table 4: Ablation study on LLM backbone replacement: comparison of generative performance across proprietary frontier and open-weight models.

Graph Generation

Graph generation builds models and algorithms that create graphs with desired structures, properties, or distributions.

See all 69 papers

ReGraph: Learning to Generate Recipe Graphs from Food Images
Aug 7, 2026Guoshan Liu, Bin Zhu, Pengkun Jiao +3Graph Generation Graph Representation Learning
Institute of Trustworthy Embodied AI, Fudan University, China and Shanghai Key Laboratory of Multimodal Embodied AI, China · Singapore Management University, Singapore
The Pragmatic Persona: Discovering LLM Persona through Bridging Inference
Apr 27, 2026Jisoo Yang, Jongwon Ryu, Minuk Ma +2Personality Modeling in Language Models Language Model Probing
Department of Artificial Intelligence, Chung-Ang University, Seoul, 06974, Republic of Korea · Department of Computer Science, University of British Columbia,2026 Vancouver, BC V6T 1Z4, Canada · Van Lang University, Ho Chi Minh City, Vietnam
When LLMs Stop Following Steps: A Diagnostic Study of Procedural Execution in Language Models
May 1, 2026Sailesh Panda, Pritam Kadasi, Abhishek Upperwal +1LLM Evaluation Instruction Following
Indian Institute of Technology Gandhinagar · Soket AI

Aug 7, 2026J/K move · Enter open · S save

ReGraph: Learning to Generate Recipe Graphs from Food Images

Guoshan Liu, Bin Zhu, Pengkun Jiao, Jingjing Chen+2

Institute of Trustworthy Embodied AI, Fudan University, China and Shanghai Key Laboratory of Multimodal Embodied AI, China · Singapore Management University, Singapore

Graph Generation Graph Representation Learning Multimodal Large Language Models

Open paper AskPDF arXiv

Recent Large Multimodal Models (LMMs) have achieved impressive performance in recipe generation from food images.However, cooking is a structured transformation process in which ingredients undergo state changes through ordered actions,while free-form recipe language leaves the corresponding entities, intermediate states, and dependencies largely implicit and entangled.A graph representation makes this procedural knowledge explicit and compositional, providing a structured basis for assessing whether model outputs encode process-level knowledge rather than merely presenting plausible textual descriptions. To address this limitation, we present ReGraph, a large-scale recipe graph dataset that represents ingredients, cooking actions, and tools as entities, uses entity attributes to describe ingredient state changes, and employs typed relations to encode manipulation targets, destinations, and procedural ordering. ReGraph further incorporates explicit Recipe Reasoning Chain-of-Thought (RR-CoT) traces, providing auxiliary supervision for procedural decomposition and structured graph generation. Building on ReGraph, we propose Recipe Graph Learning (RGL), a two-stage framework that enables LMMs to generate a plausible fine-grained cooking workflow from a food image in the form of a structured recipe graph. Under a deterministic, schema-aware matching protocol, our experiments reveal a substantial gap between text-generation quality and recoverable procedural structure: recipes produced by existing approaches achieve competitive text-generation scores yet yield limited reference-aligned entity and relation structure under the ReGraph schema. In contrast, across two representative LMM backbones, RGL consistently improves the generation of cooking entities and procedural relations, while our analysis further shows that fine-grained ingredient-state capture remains the most challenging dimension.

Abstract

Graph Generation

Explore similar work

ReGraph: Learning to Generate Recipe Graphs from Food Images

The Pragmatic Persona: Discovering LLM Persona through Bridging Inference

When LLMs Stop Following Steps: A Diagnostic Study of Procedural Execution in Language Models