# The Digital Sous Chef -- A Comparative Study on Fine-Tuning Language Models for Recipe Generation

arXiv: https://arxiv.org/abs/2508.14718 · HTML: https://mlsanity.com/paper/2508.14718/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 23 references that are in the map.
- First posted on arXiv 20 Aug 2025.

## Builds on (most cited first, 9 of 9)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://mlsanity.com/paper/1910.13461/) (2019)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [Towards a Human-like Open-Domain Chatbot](https://mlsanity.com/paper/2001.09977/) (2020)
- [Inverse Cooking: Recipe Generation from Food Images](https://mlsanity.com/paper/1812.06164/) (2018)
- [RecipeGPT: Generative Pre-training Based Cooking Recipe Generation and Evaluation System](https://mlsanity.com/paper/2003.02498/) (2020)
- [Building Language Models for Text with Named Entities](https://mlsanity.com/paper/1805.04836/) (2018)
- [Ratatouille: A tool for Novel Recipe Generation](https://mlsanity.com/paper/2206.08267/) (2022)
- [Fine-tuning Language Models for Recipe Generation: A Comparative Analysis and Benchmark Study](https://mlsanity.com/paper/2502.02028/) (2025)

## Abstract (arXiv)

We established a rigorous benchmark for text-based recipe generation, a fundamental task in natural language generation. We present a comprehensive comparative study contrasting a fine-tuned GPT-2 large (774M) model against the GPT-2 small (124M) model and traditional LSTM/RNN baselines on the 5-cuisine corpus from RecipeDB. Our key contribution is a targeted tokenization strategy that augments the vocabulary with 23 common fraction tokens and custom structural markers. This approach addresses a critical limitation of generic tokenizers by preserving essential recipe structures and precise numerical quantities, thereby enhancing domain specificity. Performance is evaluated using a comprehensive suite of seven automatic metrics spanning fluency (BLEU-4, METEOR), coherence (ROUGE-L), semantic relevance (BERTScore), and diversity. Our experiments show that the large transformer-based approach yields a >20% relative improvement in BERTScore (F1) (0.92 vs 0.72) over the best recurrent baseline, while reducing perplexity by 69.8%. We conclude with a discussion of remaining challenges, particularly regarding factual accuracy, and outline how this foundational study paves the way for integrating real-world constraints and multi-modal inputs in advanced recipe generation research.