Beyond the Final Actor: Modeling the Dual Roles of Creator and Editor for Fine-Grained LLM-Generated Text Detection

Yang Li; Qiang Sheng; Zhengjia Wang; Yehan Yang; Danding Wang; Juan Cao

cs.CLApr 6, 2026

Beyond the Final Actor: Modeling the Dual Roles of Creator and Editor for Fine-Grained LLM-Generated Text Detection

Authors: Yang Li, Qiang Sheng, Zhengjia Wang, Yehan Yang, Danding Wang, Juan Cao

Organizations: Institute of Computing Technology, Chinese Academy of Sciences · University of Chinese Academy of Sciences

Abstract

The misuse of large language models (LLMs) requires precise detection of synthetic text. Existing works mainly follow binary or ternary classification settings, which can only distinguish pure human/LLM text or collaborative text at best. This remains insufficient for the nuanced regulation, as the LLM-polished human text and humanized LLM text often trigger different policy consequences. In this paper, we explore fine-grained LLM-generated text detection under a rigorous four-class setting. To handle such complexities, we propose RACE (Rhetorical Analysis for Creator-Editor Modeling), a fine-grained detection method that characterizes the distinct signatures of creator and editor. Specifically, RACE utilizes Rhetorical Structure Theory (RST) to construct a logic graph for the creator's foundation while extracting Elementary Discourse Unit (EDU)-level features for the editor's style. Experiments show that RACE outperforms 12 baselines in identifying fine-grained types with low false alarms, offering a policy-aligned solution for LLM regulation.

Figures & tables

Reference	Target	Mean	Std
Human-Written	LLM-Polished	0.92	0.08
Human-Written	LLM-Generated	0.84	0.12
Human-Written	Humanized	0.86	0.12
LLM-Generated	Human-Written	0.84	0.12
LLM-Generated	LLM-Polished	0.89	0.09
LLM-Generated	Humanized	0.95	0.07

Table 1: Sample-level cosine similarity between rhetorical-relation frequency vectors on HART dataset.

Method	AUROC	TPR@1%FPR
Method	AUROC	Human-Written	LLM-Polished	LLM-Generated	Humanized	Avg
RoBERTa ( Solaiman et al., 2019 )	92.22	99.36	68.06	63.14	70.92	75.37
RoBERTa ${}_{\text{DANN}}$ ( Abassy et al., 2024 )	96.17	96.88	75.03	48.89	71.78	73.14
CoCo ( Liu et al., 2023 )	97.67	99.68	75.77	63.93	79.43	79.70
SeqXGPT ( Wang et al., 2023 )	89.87	98.38	15.23	14.32	31.68	39.90
DeTeCtive ( Guo et al., 2024 )	95.74	98.62	0.00	0.00	77.23	43.96

Table 2: Quantitative comparison of detection methods under the 4-class setting. For RACE, we report the results across three runs using different seeds in the format of the

\text{mean}_{\pm\text{std}}

. All values are reported in percentage (%). Bold and underlined values denote the best and second-best performance, respectively.

Method	AUROC	TPR@1%FPR
Method	AUROC	Human-Written	LLM-Polished	LLM-Generated	Humanized	Avg
RACE	97.99 ±0.13	99.04 ±0.40	83.60 ±1.61	74.18 ±0.95	75.41 ±1.03	83.06 ±0.57
w/o CL	97.73 ±0.44	98.21 ±0.19	78.07 ±1.06	69.10 ±2.66	73.43 ±1.59	79.70 ±1.12
w/o Relation	96.78 ±0.65	97.42 ±0.75	78.24 ±2.08	65.35 ±9.19	74.92 ±1.74	78.98 ±2.95
w/o RGCN	97.91 ±0.20	98.92 ±0.19	82.27 ±4.56	65.35 ±6.71	74.42 ±1.14	80.24 ±2.07
w/o Bottleneck	98.07 ±0.22	98.54 ±0.26	80.82 ±2.45	74.68 ±0.86	75.74 ±0.99	82.45 ±0.60

Table 3: Ablation results (

\text{mean}_{\pm\text{std}}

) of RACE. The Bottleneck represents the Information Bottleneck Projection in Eq. ( 2 ), the Basis represents the Basis Decomposition in Eq. ( 4 ), and w/o Relation means removing the relation types on edges and adopting vanilla GCN ( Kipf and Welling, 2017 ) .

Variant	AUROC	Avg. TPR@1%FPR
RACE (DMRST)	97.64	82.93
RACE (IsaNLP)	97.99	83.06

Table 4: Parser-robustness analysis.

Metric	CoCo	RACE
Davies-Bouldin Index ( $\downarrow$ )	0.9286	0.8042
Calinski-Harabasz Index ( $\uparrow$ )	2289.40	4333.32

Table 5: Quantitative evaluation of feature discriminability using clustering validity indices.

Domain	Method	AUROC	Avg. TPR@1%FPR
Arxiv	CoCo	93.67	55.91
	LF-Motifs	94.07	58.64
	RACE	96.61	76.28
Essay	CoCo	89.84	28.72
	LF-Motifs	91.12	47.85
	RACE	95.88	59.73

Table 6: Performance comparison under the OOD setting. We employ a Leave-One-Domain-Out protocol where the model is trained on three domains and tested on the fourth unseen domain (Column 1).

Appendix figures & tables4 assets

Supplementary material from the paper’s appendix.

Appendix

Domain	Category	Train	Val	Test	Total
Arxiv	Human-Written	700	100	200	1,000
	LLM-Polished	700	100	200	1,000
	LLM-Generated	1,229	174	352	1,755
	Humanized	172	25	48	245
Essay	Human-Written	700	100	200	1,000
	LLM-Polished	700	100	200	1,000

Appendix

Table 7: Statistics of the resplit HART dataset.

Method	Avg. TPR@5%FPR
RoBERTa	92.53
CoCo	94.13
SeqXGPT	54.41
DeTeCtive	92.82
LF-Motifs	91.90
Binoculars ${}_{\text{MLP}}$	28.00

Appendix

Table 8: Performance comparison of different detection methods evaluated using Avg. TPR@5%FPR.

Method	AUROC	TPR@1%FPR
Method	AUROC	Human-Written	LLM-Polished	LLM-Generated	Humanized	Avg
RoBERTa	98.39	99.72	72.63	70.18	71.69	78.55
CoCo	97.51	99.33	77.48	44.03	65.00	71.46
LF-Motifs	97.13	94.95	39.19	55.74	61.45	62.83
RACE (Ours)	96.59	99.13	86.25	75.20	67.00	81.90

Appendix

Table 9: Results under a stricter group-aware split that keeps all variants of the same base text in one partition.

Method	Training	Inference	Params
RoBERTa	2.119	220.8	125.2
CoCo	2.138	32.4	125.6
LF-Motifs	0.587	50.6	148.8
RACE (Ours)	1.071	90.0	128.6

Appendix

Table 10: Comparison of efficiency and model size. Training time (hours, h ), inference throughput ( samples/s ), and model size (million parameters, M ) are reported. Data preprocessing is excluded from both training time and inference throughput.

AI-Generated Text Detection

AI-generated text detection develops methods to distinguish human-written, machine-generated, and AI-edited text.

See all 170 papers

Using LLMs to Detect LLM-Generated Texts: A Cross-Generation Analysis
Sep 28, 2026Haiyue Yuan, Jie Guo, Weidong Qiu +3LLM Evaluation AI-Generated Text Detection
Institute of Cyber Security for Society (iCSS) & School of Computing, University of Kent, UK · School of Cyber Science and Engineering, Shanghai Jiao Tong University, China · School of Computer Science, University of Birmingham, UK
READER: Reasoning-Enhanced AI-Generated Text Detection
May 24, 2026Pingfan Su, Kai Ye, Shijin Gong +4AI-Generated Text Detection Large Language Model-Based Text Classification
Department of Statistics, London School of Economics and Political Science · School of Management, University of Science and Technology of China · School of Mathematics, University of Birmingham
Detecting LLM-Generated Tokens in Human--LLM Coauthored Text
Jul 23, 2026Yangjun Lu, Hongyi Zhou, Fabian Spill +3AI-Generated Text Detection Human-AI Co-Creation
School of Mathematics, University of Birmingham, Birmingham, UK · School of Statistics and Data Science, Shanghai University of Finance and Economics, Shanghai, China · Department of Statistics, The London School of Economics and Political Science, London, UK

Sep 28, 2026J/K move · Enter open · S save

Using LLMs to Detect LLM-Generated Texts: A Cross-Generation Analysis

Haiyue Yuan, Jie Guo, Weidong Qiu, Zheng Huang+2

Institute of Cyber Security for Society (iCSS) & School of Computing, University of Kent, UK · School of Cyber Science and Engineering, Shanghai Jiao Tong University, China · School of Computer Science, University of Birmingham, UK

LLM Evaluation AI-Generated Text Detection Zero-Shot Text Classification Robustness of Machine-Generated Text Detection

Open paper AskPDF arXiv

Automated detection of LLM-generated texts (LGTs) is critical, yet dedicated detectors often struggle to generalize across domains and models. While general-purpose LLMs offer flexible zero-shot authorship classification with explanatory rationale, their detection behavior, especially regarding self-detection versus cross-detection across model generations, remains poorly understood. We systematically evaluate 15 LLMs spanning three model generations as both generators and detectors. Using a benchmark of 1,000 human-written texts and 15,000 LGTs (1,000 per model), we collected over 233,000 binary classifications alongside natural-language explanations. Our results reveal that detection efficacy is primarily driven by detector capability rather than generator provenance, although outputs from newer generators remain notably harder to detect. Crucially, statistical comparisons show no systematic advantage or disadvantage for self-detection across models. Error analysis further exposes generational bias shifts: first-generation detectors under-detect LGTs (high false-negative rates), second-generation detectors over-flag human texts (high false-positive rates), and the latest models achieve balanced trade-offs. Finally, we highlight significant inconsistencies in how different LLMs apply textual cues to justify their decisions. Code: https://github.com/hyyuan/detect-llm-generated-texts.

Abstract

AI-Generated Text Detection

Explore similar work

Using LLMs to Detect LLM-Generated Texts: A Cross-Generation Analysis

READER: Reasoning-Enhanced AI-Generated Text Detection

Detecting LLM-Generated Tokens in Human--LLM Coauthored Text