Research map / Large language models / Language, safety and interpretability
Adapting Language Models for Non-Parallel Author-Stylized Rewriting
Bakhtiyar Syed, Gaurav Verma, Balaji Vasan Srinivasan, Anandhavelu Natarajan, Vasudeva Varma
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Fine-tune on the target author's corpus with DAE instead of using DAE without author-specific fine-tuning | LM + DAE · Opinosis | BLEU: 41.1 \scriptstyle{\pm 1.3} -> 43.4 \scriptstyle{\pm 1.7} | Improved |
| Fine-tune on the target author's corpus with DAE instead of using DAE without author-specific fine-tuning | LM + DAE · Opinosis | Lexical MSE: 0.33\scriptstyle{\pm 0.05} -> 0.29\scriptstyle{\pm 0.04} | Improved |
| Fine-tune on the target author's corpus with DAE instead of using DAE without author-specific fine-tuning | LM + DAE · Mark Twain | BLEU: 31.7 \scriptstyle{\pm 1.5} -> 34.4 \scriptstyle{\pm 1.8} | Improved |
| Fine-tune on the target author's corpus with DAE instead of using DAE without author-specific fine-tuning | LM + DAE · Mark Twain | Lexical MSE: 0.37\scriptstyle{\pm 0.03} -> 0.32\scriptstyle{\pm 0.03} | Improved |
| Fine-tune on the target author's corpus with DAE instead of using DAE without author-specific fine-tuning | LM + DAE · AI Wiki | BLEU: 23.7 \scriptstyle{\pm 1.6} -> 26.7 \scriptstyle{\pm 1.9} | Improved |
| Fine-tune on the target author's corpus with DAE instead of using DAE without author-specific fine-tuning | LM + DAE · AI Wiki | Lexical MSE: 0.41\scriptstyle{\pm 0.04} -> 0.34\scriptstyle{\pm 0.02} | Improved |
| Fine-tune GPT-2 on the target author's corpus | GPT-2 · Opinosis, Mark Twain, and AI Wiki | Stylistic alignment | Improved |
| Use author-specific fine-tuning rather than corpus-wide DAE fine-tuning | LM + DAE · Opinosis | ROUGE-1: 0.77 \scriptstyle{\pm 0.11} -> 0.73 \scriptstyle{\pm 0.13} | Hurt |
| Use author-specific fine-tuning rather than corpus-wide DAE fine-tuning | LM + DAE · Mark Twain | ROUGE-1: 0.68 \scriptstyle{\pm 0.14} -> 0.61 \scriptstyle{\pm 0.16} | Hurt |
| Use author-specific fine-tuning rather than corpus-wide DAE fine-tuning | LM + DAE · AI Wiki | ROUGE-1: 0.59 \scriptstyle{\pm 0.12} -> 0.54 \scriptstyle{\pm 0.13} | Hurt |
| Use non-parallel StyleLM instead of supervised stylized rewriting | Supervised stylized rewriting · Shakespeare | BLEU: 31.3 -> 33.8 | Improved |
| Use non-parallel StyleLM instead of supervised stylized rewriting | Supervised stylized rewriting · Shakespeare | Syntactic JSD: 0.17 -> 0.21 | Mixed |
Closest papers by meaning
- Non-Parallel Text Style Transfer with Self-Parallel Supervision (ICLR 2022 · similarity 0.87)
- A Probabilistic Formulation of Unsupervised Text Style Transfer (ICLR 2020 · similarity 0.87)
- RewriteLM: An Instruction-Tuned Large Language Model for Text Rewriting (AAAI 2024 · similarity 0.85)
- Imitate Before Detect: Aligning Machine Stylistic Preference for Machine-Revised Text Detection (AAAI 2025 · similarity 0.84)
- ParaGuide: Guided Diffusion Paraphrasers for Plug-and-Play Textual Style Transfer (AAAI 2024 · similarity 0.84)
- Raidar: geneRative AI Detection viA Rewriting (ICLR 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
