Star on GitHubStar3.2k stars

2026 · Preprint

BUSTA: Teaching Small Language Models to Reason About Binary Execution

BUSTA trains small language models to infer input register values from an x86-64 basic block and desired output register state. It evaluates knowable-register prediction accuracy against base and frontier models.

Scope and limitations

The evaluation focuses on inverse execution of x86-64 basic blocks and register values.

Why it fits this collection

Inferring the inputs that produce a binary block’s output is reasoning about existing binary code behavior.

Cite

@misc{paper-busta,
  title = {{BUSTA: Teaching Small Language Models to Reason About Binary Execution}},
  author = {{Ritika Poddar} and {Aidan Fennelly} and {Deirdre Murphy} and {Eric Wolos} and {Corre L. Steele} and {Michael Doyle}},
  year = {2026},
  url = {https://arxiv.org/abs/2610.36288},
  howpublished = {arXiv:2610.36288},
  note = {Preprint},
  eprint = {2610.36288},
  archivePrefix = {arXiv},
  month = oct
}

Read in a terminal: hf papers read 2610.36288

Publication and review details

Venue
arXiv
Status
Preprint
Publication / posting date
October 26, 2026
Date source
First arXiv submission date.
First arXiv submission
October 26, 2026
arXiv
2610.36288
Reviewed arXiv version
v1

The date and venue come from the arXiv record.

Sources reviewed

Reviewed against the paper and its primary sources. Cite the original authors using the linked paper or BibTeX.

Report an outdated entry Suggest a translation fix

Primary sources

  1. https://arxiv.org/abs/2610.36288
↑↓ move↵ open⌘↵ new tab