Caroline Lemieux
@cestlemieux.bsky.social
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.
Was already on my reading list, gives me more incentive to read it :P
November 28, 2024 at 7:52 PM
Was already on my reading list, gives me more incentive to read it :P
Reposted by Caroline Lemieux
🧠 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/
November 26, 2024 at 4:09 PM
🧠 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/