FormulaSPIN: Self-Play Fine-Tuning for Natural Language to Spreadsheet Formula Generation
Existing approaches rely on static supervised data, which quickly saturates on limited annotations. In this paper, we introduce FORMULASPIN, a self-play framework that breaks the ceiling of supervised fine-tuning by enabling iterative self-improvement without any additional data. Vanilla SPIN fails on this task: it uniformly penalizes every non-matching output, so execution-equivalent alternatives are punished as negatives in one example while serving as ground truth in another, producing contradictory gradients.
- ▪Existing approaches rely on static supervised data, which quickly saturates on limited annotations.
- ▪In this paper, we introduce FORMULASPIN, a self-play framework that breaks the ceiling of supervised fine-tuning by enabling iterative self-improvement without any additional data.
- ▪Vanilla SPIN fails on this task: it uniformly penalizes every non-matching output, so execution-equivalent alternatives are punished as negatives in one example while serving as ground truth in another, producing contradictory gradients.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2607.19354 (cs) [Submitted on 21 May 2026] Title:FormulaSPIN: Self-Play Fine-Tuning for Natural Language to Spreadsheet Formula Generation Authors:Cy Xie View a PDF of the paper titled FormulaSPIN: Self-Play Fine-Tuning for Natural Language to Spreadsheet Formula Generation, by Cy Xie View PDF HTML (experimental) Abstract:Spreadsheet applications are used by hundreds of millions worldwide, yet writing formulas remains a significant barrier. Existing approaches rely on static supervised data, which quickly saturates on limited annotations. In this paper, we introduce FORMULASPIN, a self-play framework that breaks the ceiling of supervised fine-tuning by enabling iterative self-improvement without any additional data.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv.org.
Discussion
0 commentsMore from arXiv.org
