Lightnews — Scholar-powered news

Caroline Lemieux

@cestlemieux.bsky.social

370 followers 99 following 9 posts

now: Assistant Professing™ in Software Practices Lab at UBC. was: postdoc MSR NYC, phd UC Berkeley. also at https://mastodon.acm.org/@cestlemieux. she/her.

mastodon.acm.org

Posts Media Videos Starter Packs

Caroline Lemieux @cestlemieux.bsky.social · 1d

The UBC Software Practices Lab is heading to #icfpsplash25! 4 ICFP/OOPSLA talks, 1 SPLASH-E, 5 talks at associated workshops... check it out: www.cs.ubc.ca/news/2025/10...

UBC Computer Science makes waves at programming language conference ICFP/SPLASH

www.cs.ubc.ca

1 6

Caroline Lemieux @cestlemieux.bsky.social · Jun 25

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...

Release Pynguin 0.41.0 · se2p/pynguin

Fix subject_properties aren't registered when running Pynguin on an imported module Update documentation (Codestyle, Code Overview) Add LLM-Agent guidelines Add PynguinML mode: Parsing and test gen...

github.com

Caroline Lemieux @cestlemieux.bsky.social · Jun 9

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:

Rohan Padhye @rohan.padhye.org · Jun 6

Just Accepted to ACM TOSEM!

The "Havoc Paradox" is about the relationship between byte-level fuzzer mutations and their effect on the inputs produced by generators for structured strings (e.g. XML/SQL). Can disruptive mutations be controlled? Should they be? Find out.

📄 dl.acm.org/doi/pdf/10.1...

Caroline Lemieux @cestlemieux.bsky.social · Jun 4

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

Caroline Lemieux @cestlemieux.bsky.social · May 27

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...

Caroline Lemieux @cestlemieux.bsky.social · May 5

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!

Caroline Lemieux @cestlemieux.bsky.social · Apr 4

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 :).

1 8

Caroline Lemieux @cestlemieux.bsky.social · Mar 21

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

FUZZING'25 Workshop @ ISSTA

The 4th International Fuzzing Workshop (FUZZING) 2025 welcomes all researchers, scientists, engineers and practitioners to present their latest research findings, empirical analyses, t...

fuzzingworkshop.github.io

2 6 10

Caroline Lemieux @cestlemieux.bsky.social · Nov 28

Was already on my reading list, gives me more incentive to read it :P

Reposted by Caroline Lemieux

Dr. Claire Le Goues @clegoues.bsky.social · Nov 26

🧠 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/

1 1 5