2026 · Conference accepted
CHISEL-ing Back Source Code with AI-enabled Iterative Recovery
CHISEL refines Ghidra pseudocode using compiler diagnostics and type-aware differential fuzzing against the original function. It retains counterexamples and candidate history; existing I/O pairs are held out for evaluation rather than supplied to the recovery loop.
Scope and limitations
Evaluation covers 120 selected ExeBench functions, x86-64 O0–O3, stripped/unstripped variants and a fixed Gemma4:31b model. The feedback oracle still falsely accepts candidates; finite fuzzing and held-out tests do not establish equivalence. Signature recovery and possible benchmark exposure remain limitations.
Why it fits this collection
Adds test-suite-free in-loop refinement while explicitly separating oracle acceptance from verified behavior.
Publication and review details
- Venue
- SURE @ ACM CCS 2026
- Status
- Conference accepted
- Publication / posting date
- 2026 (exact date not verified)
- Date source
- Acceptance is reported in the author arXiv record and publications page; proceedings publication is not verified.
- First arXiv submission
- August 28, 2026
- arXiv
- 2608.27981
- Reviewed arXiv version
- v1
Author sources report acceptance to SURE 2026; the manuscript DOI remains a placeholder, so no proceedings publication date or DOI is assigned. The paper says selection requires at least 200 tests, while the pinned README says at least 100 held-out I/O pairs; exact evaluated artifact alignment remains unresolved. The article is CC BY 4.0, but the inspected root/README did not establish a code/data license. Nothing was run.
Sources reviewed
Source review only; experiments were not rerun for this collection. Cite the original authors using the linked paper or BibTeX.