CodaMOSA was built on Pynguin version 0.18 .... since then, Pynguin has been very much improved. Pynguin version 0.41.0 includes the LLMMOSA Algorithm, which integrates core parts of CodaMOSA into modern Pynguin! Many kudos to Pynguin contributors for this. github.com/se2p/pynguin...

A common follow-up, whenever we bring up parametric generators is "but can't those byte-level mutations result in a totally different generated input?" ....they do, but our experiments found those more destructive mutations generally lead to higher coverage. Our students fully explored this in:

after years of fuzzing libxml2 I am happy to announce I have now actually used xmllint

I was today days old when I learned that IEEETran format has special figure captions overwritten by \usepackage{subcaption}; instead one should use \usepackage[caption=false]{subfig}. tex.stackexchange.com/questions/30...

Back from #ICSE25! I'm looking forward to reading "No Harness, No Problem: Oracle-guided Harnessing for Auto-generating C API Fuzzing Harnesses", by @gabriel-sherman.bsky.social and @snagycs.bsky.social.(users.cs.utah.edu/~snagy/paper...). Nice progress in fuzz driver generation!

Finally added a little logo for QuAC (our attribute-based python type inference tool, paper: doi.org/10.1145/3689... + repo: github.com/jifengwu2k/q...) to my website :).

There's still time to submit to FUZZING'25! This year, we're accepting both the (now classic) registered reports _and_ new short papers (fuzzing nuggets). Deadline is now March 26th! fuzzingworkshop.github.io

🧠 Older models memorize more: Models like CodeGen, and CodeLlama show significantly higher leakage on Defects4J than newer models (e.g., Llama 3.1). They often reproduce patches verbatim, to the point that it’s weird (including comments!!) 🔥 3/