Robert Feldt

Keynote slides

The AIST workshop keynote by @rfeldt.bsky.social is about to begin. Join us in room B at 14:00.

AIST 2025 - 5th International Workshop on Artificial Intelligence in Software Testing - ICST 2025 The integration of AI techniques in the domain of software testing represents a promising frontier, one that is still at the dawn of its potential. Over the past few years, software developers have wi...

In two weeks (April 1st), the 5th edition of the Artificial Intelligence in Software Testing (#AIST) workshop will be held, an event colocated with the ICST conference. In this thread, we'll tell you about the program Aitor Arrieta and I are preparing:

conf.researchr.org/home/icst-20...

Capturing Semantic Flow of ML-based Systems ML-based systems are software systems that incorporates machine learning components such as Deep Neural Networks (DNNs) or Large Language Models (LLMs). While such systems enable advanced features suc...

Here's something that I am genuinely excited about. We propose a graph-like representation for executions of LLM agents called semantic flow. Intuitively speaking, it connects clusters of latent vectors which are embeddings of LLM outputs. Will appear at FSE 2025 IVR track.

arxiv.org/abs/2503.10310

This is so tragically apropos

It's that time of year where I post this one

Winter 2025 - Researcher positions in AI and machine learning — FCAI

Postdoc and doctoral student positions in developing Bayesian methods! The positions are funded by Finnish Center for Artificial Intelligence FCAI and there are many other topics, too, but if you specify me as the preferred supervisor then it's going to be Bayesian. fcai.fi/winter-2025-...

From Opinion to Evidence: An Empirical Approach to Software Engineering. The recording of my inaugural lecture at Uni Bayreuth is now available online: youtu.be/58OvavOQO0w

I’m excited to share that our paper "An Analysis of #LLM Fine-Tuning & Few-Shot Learning for Flaky Test Detection & Classification" has been accepted for publication at the Int’l Conf on Software Testing, Verification & Validation (#ICST2025). This paper is first-authored by @riddhimore.bsky.social

+1

Practically investigating "willingness to pay for it" these days. More ROI studies needed - I'll try to contribute some going forward.

ESEM was the first international conference where I published, and for a decade now, I have studied and used the guidelines for empirical software engineering developed by the ISERN community. In a way, ISERN was my first academic community, and today I am honored and happy to join them as a member!

Ever wondered how language models predict if an issue is a bug or not? We found that they are quite robust and that they use contextual and relevant information to make these decisions!

And now we get to discuss this with you @icseconf.bsky.social at the J1 track!

Paper: doi.org/10.1007/s106...

Congrats to my student Fabian Pena for getting his paper accepted at the doctoral symposium of @icseconf.bsky.social!

His PhD focuses on understanding on software engineering knowledge within language models and it will be exiting for to hear the feedback from the community for this topic.

Working with language models? Want to avoid getting sued? But you have no idea what might get you in trouble?

We had the same problem!

We teamed up with legal experts from our university to shed light on data protection and license issues.

Early access in IEEE Software: doi.org/10.1109/MS.2...

Hunting bugs: Towards an automated approach to identifying which change caused a bug through regression testing - Empirical Software Engineering Context Finding code changes that introduced bugs is important both for practitioners and researchers, but doing it precisely is a manual, effort-intensive process. The perfect test method is a theore...

Hunting bugs: Towards an automated approach to identifying which change caused a bug through regression testing by Maes Bermejo, yt, Gallego Carrillo, Gortázar Bellas, Robles, Gonzalez-Barahona accepted to be presented at ICSE 2025 (Journal First). link.springer.com/article/10.1...

SE researchers try advanced ML for "everything" (and conferences encourage it by demanding "novelty") but few investigate that complexity-vs-accuracy ROI for practitioners.

So maybe this is a case where advanced ML/non-linear modelling might not be the right/best ROI. I suspect there are many such SE areas/tasks...

Ranking approaches for similarity-based web element location GUI-based tests for web applications are frequently broken by fragility, i.e. regression tests fail due to changing properties of the web elements. Th…

New paper in JSS:
We tried Learning-to-Rank algorithms to improve web element localisation. Found (small) benefits vs state-of-the-art. Accuracy typically in 87-95% range for 40 web apps. Linear/simpler models show quite some promise though.

www.sciencedirect.com/science/arti...

Senior Lecturer of Trustworthy AI | University College Cork Learn, Study and Research in UCC, Ireland's first 5 star university. Our tradition of independent thinking will prepare you for the world and the workplace in a vibrant, modern, green campus.

My department is hiring! Talk to me if you want to know more.

Senior Lecturer (=Assoc. Prof) in Trustworthy AI: www.ucc.ie/en/futures/a...

Lecturer (=Assist Prof) in AI-powered Augmented & Virtual Reality
www.ucc.ie/en/futures/a...

Lecture in Mobile / Edge Computing: www.ucc.ie/en/futures/a...

ICSE 2025 ICSE, the IEEE/ACM International Conference on Software Engineering, is the premier software engineering conference. It will be held April 27-May 3 2025 in Ottawa. Core conference days will be Wednesd...

ICSE, the IEEE/ACM International Conference on Software Engineering, is the premier software engineering conference. It will be held April 27-May 3 2025 in Ottawa. Core conference days will be Wednesday April 30 to Friday May 2.

conf.researchr.org/home/icse-2025

On my way back from a fantastic 10 days of research and new ideas with @drsyoo.bsky.social and his whole group @coinse.bsky.social
Big thanks to you all for hosting me and for all the good times/parties/brainstorming 🙏🏻
See you soon again 👍🏻👍🏻👌🏼

Delighted to be co-organising the inaugural CauSE 2025 workshop (Causal Methods in Software Engineering), co-located with FSE in Trondheim.

We welcome papers on all things causality and SE! Deadline Feb. 25th.

Spread the word!

causality-software-engineering.github.io/cause-worksh...

Very exciting indeed.

For sure, that’s how we roll 😆😂👌🏼

Having an exciting afternoon with @rfeldt.bsky.social here in KAIST, cooking up ideas after ideas until they get really wacky... this is how you go about it, right?

Honolulu

ESEM, the International Symposium on Empirical Software Engineering and Measurement, is the premier empirical software engineering and measurement conference.

ESEM2025 : September, 28 - October 3, 2025 in Honolulu, Hawaii, US
bit.ly/49czfAz

Another really interesting quote from Anthropic’s Amodei:
”The farther a skill is from the people building AI the longer it takes before AI disrupts it”

I need to check out this guy more…

Really interesting quote from Anthropic CEO Amodei:
”Talent density beats talent mass”

(Even if you have more top talent they will be slowed down and not as inspired to be their best if they mainly see less talented/driven people when looking around in the team/org)

Feels like a fresh start. Let’s see what happens over time… 😬

On my way back to visit @drsyoo.bsky.social and his excellent group again! I’m so happy and looking forward to 10 days of research, brainstorming and fun! 👍🏻🤩

On my way there Helsinki greets with snow:

TIL about hijacked journals, predatory journals impersonating legitimate ones. retractionwatch.com/the-retracti... “Conference clones” existed for many years, but I have never heard about such journals…