4 ms·
Huh, I wonder if LLMs are good at fuzzing in general.
by UltimateEdge 4y ago
Huh, I wonder if LLMs are good at fuzzing in general.
- summarity 4y agoAlso did some early experiments but didn't go very far. If someone wants to investigate: use structural code analysis tools (parser-parser-generators) like Comby[1] to factor existing code, letting GPT remix it instead of generating test cases all by itself. [1] https://comby.dev/blog/2022/04/11/comby-decomposer-compiler-fuzzing https://comby.dev/blog/2022/04/11/comby-decomposer-compiler-...