# Does reverse engineering need a bigger model? General and specialist models in Touhou reconstruction

> An essay on REA’s site by N0zoM1z0 (13 October 2026) uses dated Touhou reconstruction commits to ask whether exact function matching needs a larger general model, a compiler-specialist model or better retrieval. It walks through TH09, TH095, TH04 and TH08 cases where later GPT-6-series sessions closed stubborn matches, and the TH08 audit that found accepted functions with wrong floating constants. Specialist and reinforcement-learning comparisons are proposed as future experiments; no trained specialist was evaluated.

- Type: Case study
- Targets: Binary · Game
- Category: Official REA: case studies
- Relationship: Official REA case study
- Sources reviewed: 2026-10-19
- Added to the catalog: 2026-11-02
- Original resource: <https://rea.tools/blog/reverse-engineering-models/>
- Primary source: <https://github.com/morluto/rea/blob/e8da7e1edd075feec0e6beedec93d5645fed8125/website/evidence/reverse-engineering-models.md>
- Catalog page: <https://awesomerea.org/resources/rea-bigger-model-essay-case/>
- Record JSON: <https://awesomerea.org/resources/rea-bigger-model-essay-case/record.json>
- Chinese version: [简体中文](https://awesomerea.org/zh/resources/rea-bigger-model-essay-case/index.md)

## Requirements & limits

- **Prerequisites:** The essay and its notes supply no game files, compilers or builds; each linked Touhou repository documents its own requirements.
- **Scope:** The essay is dated 13 October 2026. Its examples are function-level comparisons recorded in linked TH09, TH095, TH04 and TH08 commits.
- **Limits:** Later sessions inherited earlier source, diagnostics and failed experiments, so the before-and-after closures are not controlled estimates of model-size effects. The notes treat gpt-web as a changing label that does not identify a run’s model, per-run reasoning effort and budgets are not uniformly recorded, and the GPT-6-Luna-max closure is presented as a discovery across model tiers, not a measured smaller-model advantage.
- **Testing evidence:** The TH08 example comes from commit a393f400 (2 September), which target-checks every VC7 floating-literal relocation; the follow-up audits 1,548 relocations and corrects 12 stale rows in five accepted units. The essay’s concern that a reward built on the old check could reinforce wrong source is stated as an inference: the notes record no actual RL exploit or trained model doing so.
- **License:** Game rights stay with their owners.

## Primary sources

1. <https://github.com/morluto/rea/blob/e8da7e1edd075feec0e6beedec93d5645fed8125/website/evidence/reverse-engineering-models.md>
2. <https://github.com/morluto/rea/blob/e8da7e1edd075feec0e6beedec93d5645fed8125/website/public/blog/reverse-engineering-models/index.html>
3. <https://github.com/N0zoM1z0/th08/commit/a393f400786e4cc6cbb1000af0dad3276da12c78>

## More in Games

- [FH2MP: reconstructed Forza Horizon 2 online services](https://awesomerea.org/resources/fh2mp-case/index.md) · Case study
- [game-and-watch-patch: Nintendo Game & Watch custom firmware](https://awesomerea.org/resources/game-and-watch-patch-cfw-case/index.md) · Case study
- [GTA Mode: reverse-engineering SimCity 3000 Unlimited for a native plugin](https://awesomerea.org/resources/simcity-3000-gta-mode-case/index.md) · Case study
- [inFAMOUS Second Son native PC port](https://awesomerea.org/resources/infamous-second-son-pc-port-case/index.md) · Case study
- [Left 4 Dead 2: recovering the AI Director's behavior](https://awesomerea.org/resources/l4d2-director-research-case/index.md) · Case study
- [MTG Arena Collection Export: reconstructing the Frontdoor collection request](https://awesomerea.org/resources/mtg-arena-collection-export-case/index.md) · Case study

## About this entry

The description was reviewed against primary sources.
Original editorial content: MIT. Linked projects, papers and assets retain their own rights and licenses.