iiinova — where your digital self keeps evolving

Lukas Pirch

@lpirch.bsky.social

67 followers 55 following 3 posts

PhD at BIFOLD, TU-Berlin • Vulnerability Discovery & Graph-based Machine Learning • 🎹🎸

Posts Replies Media

Reposted by Lukas Pirch

anil oza @aniloza.bsky.social · 14/05/2026

wow — the preprint host, arxiv, is banning authors for a year if they submit papers with hallucinated citations 🤖

Twitter thread from Thomas Dietterich, reading "Attention
@arxiv
authors: Our Code of Conduct states that by signing your name as an author of a paper, each author takes full responsibility for all its contents, irrespective of how the contents were generated. 1/
3:03 PM · May 14, 2026
·
66.7K
Views
Relevant
View quotes

Thomas G. Dietterich
@tdietterich
·
1h
If generative AI tools generate inappropriate language, plagiarized content, biased content, errors, mistakes, incorrect references, or misleading content, and that output is included in scientific works, it is the responsibility of the author(s). 2/
Thomas G. Dietterich
@tdietterich
·
1h
We have recently clarified our penalties for this. If a submission contains incontrovertible evidence that the authors did not check the results of LLM generation, this means we can't trust anything in the paper. 3/
Thomas G. Dietterich
@tdietterich
·
1h
The penalty is a 1-year ban from arXiv followed by the requirement that subsequent arXiv submissions must first be accepted at a reputable peer-reviewed venue. 4/
Thomas G. Dietterich
@tdietterich
·
1h
Examples of incontrovertible evidence: hallucinated references, meta-comments from the LLM ("here is a 200 word summary; would you like me to make any changes?"; "the data in this table is illustrative, fill it in with the real numbers from your experiments") end/"

9458121915

Reposted by Lukas Pirch

Konrad Rieck 🌈 @rieck.mlsec.org · 18/04/2026

Yesterday, Lukas presented at #ICSE2026 results I didn't believe when he first showed them to me. A classifier based on simple code metrics matches frontier LLMs for vulnerability discovery in performance, as long as they are not agentic. 🧵 (1/4)

Paper: LLM-based Vulnerability Discovery through the Lens of Code Metrics

1112

Lukas Pirch @lpirch.bsky.social · 18/04/2026

Code Metrics perform as well as LLM-based classifiers for source code vulnerabilities! I'm glad the community didn't shoot the messenger when I presented our recent work at ICSE'26. preprint: mlsec.org/docs/2026-ic... Shout out to Felix, Erik, Jonas, @eisenhofer.bsky.social and @rieck.mlsec.org

Lukas Pirch at ICSE'26 next to his title slide

041

Reposted by Lukas Pirch

Explainable AI Berlin @xai-berlin.bsky.social · 28/11/2025

Manipulating Feature Visualizations with Gradient Slingshots @dilya.bsky.social Marina MC Höhne, Alexander Warnecke @lpirch.bsky.social Klaus-Robert Müller @rieck.mlsec.org @slapuschkin.bsky.social @kirillbykov.bsky.social 👇

143

Reposted by Lukas Pirch

Konrad Rieck 🌈 @rieck.mlsec.org · 15/10/2025

AI predicts rain. We predict trouble! Today, Erik presents a novel attack on Google's latest AI weather model at #CCS2025. By changing only 0.1% of the observations, the attack can fabricate or suppress the prediction of extreme events, from hurricanes 🌀 to heat waves 🔥 1/4 @bifold.berlin

Predicted temperature at the peak of the European Heat Wave 2006

2215

Lukas Pirch @lpirch.bsky.social · 30/04/2025

Photogenic as always 😂

040

Reposted by Lukas Pirch

Konrad Rieck 🌈 @rieck.mlsec.org · 10/04/2025

Great to be at @satml.org with several members of my team from @bifold.berlin and @tuberlin.bsky.social. We are having a blast with exciting discussions and talks on trustworthy AI! #SaTML25

1153

Lukas Pirch @lpirch.bsky.social · 09/04/2025

initial commit. add README and a pic of us driving to #SaTML25 this morning 🚲☀️

040