Reverse-engineering data for AI labs
RL environments and eval sets built on real compiled binaries. Every task ships with a grader you can rerun, a reference solution and fresh instances on demand.
RE-Lab: verified reverse-engineering tasks
Real binaries: Seven CPU architectures plus Windows
Real compiled programs on x86-64, i686, AArch64, ARM32, RISC-V, MIPS and PPC64LE, plus Windows PE. Each task asks a question the agent has to answer from the binary itself.
Exact graders: Deterministic grading
Hidden test cases, scored the same way on every run. Every task ships with a reference solution and control checks that show empty and random answers fail. Difficulty is measured on named models, and everything runs offline.
Packs and pilots: Ready-made or built to your spec
Start with a ready-made evaluation pack, or a scoped pilot of task families built to your spec, delivered in batches against an objective acceptance test. If a set doesn't fit your work, we build a replacement within a week.
Contact: Write to us
Tell us what you're training or evaluating, and we reply in writing with a scoped proposal.
© 2026 RE-Lab · Rovaniemi, Finland. Interoperability and software understanding only; no malware, exploit or evasion tasks.