LOADING
Code generation is one of the most promising applications of large language models (LLMs), offering substantial productivity boosts for developers. However, this benefit is tempered by serious concerns surrounding the correctness and security of the generated code - especially outside the happy path.
Code generation is one of the most promising applications of large language models (LLMs), offering substantial productivity boosts for developers. However, this benefit is tempered by serious concerns surrounding the correctness and security of the generated code - especially outside the happy path.
Several solutions have been put forward to mitigate this, ranging from agentic code review to exhaustive test generation to formal verification.
In this talk, I present evidence from our research that these approaches are not currently sufficient to mitigate those concerns and instead suggest an alternative, more viable path forward based on the Scala type system.
I conclude by offering an insight from the perspective of LLM developers on why this approach is working and how even more secure coding LLMs can be developed, thanks to Scala features.
Drawing from multiple Scala LLM workshops we conducted this past year, I will share insights to significantly enhance your AI experience.
This talk presents McCCT, a new concurrency testing tool developed at KTH by the speakers in the context of an ongoing research project.
Don't miss out on this opportunity to connect with Scalar community and create lasting memories!
Scala Native can interact with C code and libraries, greatly expanding the library ecosystem beyond pure Scala offerings. Let's see the low level and high level tools that make it possible, talk through challenges of encoding various C concepts in Scala, and demonstrate what popular C libraries look like when used alongside idiomatic Scala code.
Protobuf is commonly associated with code generation. However, in large projects with tens of thousands of message definitions, this approach can lead to an overwhelming amount of generated code. In this talk, I’ll share my journey in search of a different approach to this problem.
