Chun Feng

Research

Papers, deep dives, and experiments on LLM-assisted reverse engineering and the emerging practice of Vibe RE.

An elegant, matte-black notebook open on a wooden desk, its pages filled with carefully drawn flowcharts, abstract callgraph diagrams, and symbolic arrows representing control flow, all non-textual. Beside it, a tablet displays a stylized neural network diagram overlaying a blurred decompiler-style code structure. The environment is a quiet study with dark wood shelves and softly glowing ambient light strips along the walls. Warm, indirect evening lighting casts gentle shadows and a golden sheen on the paper edges. Photographic realism, shot from a three-quarter overhead angle with moderate depth of field, creates a contemplative, sophisticated mood, illustrating the blend of traditional reverse engineering analysis and modern LLM-assisted reasoning for long-form writing and research.
A close-up of a polished black PCB resting on a clean white desk, with a transparent acrylic overlay etched with abstract patterns hinting at neural network connections merging into traces and vias. In the background, slightly blurred, a monitor shows a high-level visualization of data flow analysis without any readable text. Cool, directional studio lighting from the left creates crisp highlights on the copper traces and subtle reflections in the acrylic. The mood is analytical and precise, with a modern, high-tech aesthetic. Photographic realism, captured from a slightly elevated macro angle, emphasizes fine textures and details, framing the board along the rule of thirds to suggest the meeting of hardware reverse engineering and LLM-powered insight.

LLMs at the Disassembly Layer

I study how large language models change the practice of reverse engineering: assisting disassembly, reasoning about binaries, and mapping opaque systems. This site collects selected research, experiments, and notes from ongoing work exploring that intersection.

Read research

Logbook

Chronological research notes, experiments, and technical reverse engineering logs.

An elegant, matte-black notebook open on a wooden desk, its pages filled with carefully drawn flowcharts, abstract callgraph diagrams, and symbolic arrows representing control flow, all non-textual. Beside it, a tablet displays a stylized neural network diagram overlaying a blurred decompiler-style code structure. The environment is a quiet study with dark wood shelves and softly glowing ambient light strips along the walls. Warm, indirect evening lighting casts gentle shadows and a golden sheen on the paper edges. Photographic realism, shot from a three-quarter overhead angle with moderate depth of field, creates a contemplative, sophisticated mood, illustrating the blend of traditional reverse engineering analysis and modern LLM-assisted reasoning for long-form writing and research.

Artifacts, Diagrams, and Data

Supplemental material that supports the main research: standalone diagrams, datasets, annotated traces, and external references that sit between a paper and a full project. Use this space as a technical appendix to the rest of the site.

View artifacts