# Stage-Supervised Latent Reasoning for Single-Shot JavaScript Deobfuscation

> Uses intermediate Webcrack rewrites to supervise Coconut-style latent reasoning for JavaScript deobfuscation. Training exposes staged code transformations; inference receives obfuscated code and generates a cleaned program in one shot.

- Authors: Rong Feng; Suman Saha
- Venue: arXiv
- Status: Preprint
- Publication / posting date: September 22, 2026
- Date source: First arXiv submission; no formal proceedings publication was verified.
- First arXiv submission: September 22, 2026
- arXiv: 2609.27058
- Reviewed arXiv version: v1
- DOI: 10.48550/arxiv.2609.27058
- Sources reviewed: 2026-10-20
- Added to the catalog: 2026-10-20
- Original paper: <https://arxiv.org/abs/2609.27058>
- PDF: <https://arxiv.org/pdf/2609.27058v1>
- BibTeX: <https://awesomerea.org/papers/citations/paper-stage-supervised-js-deobfuscation.bib>
- Catalog page: <https://awesomerea.org/papers/paper-stage-supervised-js-deobfuscation/>
- Record JSON: <https://awesomerea.org/papers/paper-stage-supervised-js-deobfuscation/record.json>
- Chinese version: [简体中文](https://awesomerea.org/zh/papers/paper-stage-supervised-js-deobfuscation/index.md)

## Scope and limitations

A work-in-progress study of one 3B backbone: 100 programs sampled from the 300 smallest benchmark inputs, split 60/20/20. Of 20 test outputs, 10 parsed and 8 matched teacher-derived outputs; the reported 80% is conditional on valid syntax. Semantic checks include manual inspection; stronger baselines and stage ablations remain future work.

## Why it fits this collection

Explores how deterministic recovery stages can supervise learned code deobfuscation.

## Publication and review notes

Reviewed September 22 v1, explicitly labeled WIP, with a registered arXiv DOI. Article license is CC BY 4.0. Webcrack is a cited upstream supervision tool, not this paper’s implementation release; no dedicated author code, weights or training-artifact license was verified.

## Primary sources

1. <https://arxiv.org/abs/2609.27058v1>
2. <https://arxiv.org/html/2609.27058v1>
3. <https://api.datacite.org/dois/10.48550/arxiv.2609.27058>

## About this entry

Reviewed against the paper and its primary sources. Cite the original authors using the linked paper or BibTeX.
Original editorial content: MIT. Linked projects, papers and assets retain their own rights and licenses.