I hadn’t seen a whitepaper for a while and I wondered, are folks diving into the details of advances in Large Language Model (LLM) tech via whitepapers? If so, where do you get your LLM related whitepapers from?
And feel free to comment on this particular whitepaper if you like. Have you seen it before reading this post? What did you make of it?
I don’t read white papers much, but as I work in Fin-Tech I was exploring the use cases of LLM in finance and recently came across the below white paper, which focuses on the same:
I usually find such white papers in blogs and articles on different platforms.
At some point in your career, you realize that “there are no new or interesting topics” in testing or automation. (Especially when you have not transitioned to the leadership role). Here comes the whitepapers.
How to search for an interesting whitepaper: find one, go to references, and find even more papers on the topic!
To search for a free whitepaper - use Google Foo and add ‘filetype:pdf’ to your search phrase
Make a list of papers and read it one by one
Some papers are more “general” one - so you can skip part of it
Some papers contain math - so you might spend some time getting through it
Make notes from papers! (There are a bunch of note-taking techniques)
Papers on LLMs
Papers that I added to my ToDo (after reading the paper I mentioned on TWiT session):
Testing Web-Enabled Simulation at Scale Using Metamorphic Testing
Automated Unit Test Improvement using Large Language Models at Meta
Software Testing Research Challenges: An Industrial Perspective
Assured LLM-Based Software Engineering
Observation-based unit test generation at Meta
The Oracle Problem in Software Testing: A Survey
What It Would Take to Use Mutation Testing in Industry—A Study at Facebook
ChatUniTest: A Framework for LLM-Based Test Generation
Mutation-guided LLM-based Test Generation at Meta
I plan to start making reviews on papers in my blog. Because I am collecting a lot of information and not sharing it with the world. This thing should be fixed
I haven’t, yet, started reading on LLM-Based Test Generation. I would appreciate any feedback on the matter (articles, whitepapers…) has it’s on my top of things to lookout this year.
About META’s Mutation-Guided LLM-Based Test Generation paper, it’s a hard and long read. I’ve found an easier article that gives the high level view:
On the Biology of a Large Language Model uses that methodology to investigate Claude 3.5 Haiku in a bunch of different ways. Multilingual Circuits for example shows that the same prompt in three different languages uses similar circuits for each one, hinting at an intriguing level of generalization.