HiLDe : Intentional Code Generation via Human-in-the-Loop Decoding
Emmanuel Anaya González*, Raven Rothkopf*, Sorin Lerner, Nadia Polikarpova · 2025
Read HiLDe : Intentional Code Generation via Human-in-the-Loop Decoding by Emmanuel Anaya González*, Raven Rothkopf*, Sorin Lerner, Nadia Polikarpova (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-22906.txt.
Description
Emmanuel Anaya González and co-authors worry that AI coding assistants make programmers passive, and they build a tool to keep humans in charge. Their system, HiLDe, lets a programmer watch and steer a language model while it writes code, rather than accepting text. The assistant highlights moments when the model makes a critical choice and offers local alternatives the user can pick, so intent guides generation at each fork.
Eighteen programmers used HiLDe and a standard assistant on security-related tasks, each trying both. Code written with the steering tool carried far fewer security flaws, and it matched the programmers' goals more closely. The authors argue the risk: in fields like software security, accepting code on autopilot can plant weaknesses that are costly to find. Their fix is not a smarter model but a better conversation with the model.
Catalog record
| CREATOR | Emmanuel Anaya González*, Raven Rothkopf*, Sorin Lerner, Nadia Polikarpova |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-22906