Patrick Knab

PhD Candidate - ML - Technical University of Clausthal

[Patrick Knab @p_knab]

@asmith4981 @ajassy Agreed. It also skips why the astrophage all died during the return trip, so the centrifuge solution feels like it comes out of nowhere.

[Patrick Knab @p_knab]

@giangnguyen2412 Both on iPhone and iPad

[Patrick Knab @p_knab]

@DarkPoolKnight @icmlconf I was 30min late and already sold out.

[Patrick Knab @p_knab]

@icmlconf Are reserved spots also available for workshop-paper authors or volunteers? In particular for workshop authors, travelling to Seoul for only two days is difficult if registration is already full.

[Patrick Knab @p_knab]

From these graphs, we derive zero-shot VLM tasks, including phrase grounding, HOI recognition, referring, activity recognition, relation extraction, and temporal VQA. You can also sample more instances around hard tasks and reuse the labeling + graph recipe for your own dataset.

[Patrick Knab @p_knab]

Check out our new preprint, dataset, and benchmark BARISTA if you are a coffee lover like we are and interested in VLM benchmarking and debugging temporal & spatial reasoning in egocentric video.

Preprint: https://t.co/rp9fhGKQ0j
Dataset: https://t.co/oGxUfcPoc4

[Patrick Knab @p_knab]

BARISTA provides densely annotated coffee-preparation videos with per-frame scene graphs, object identities, masks, boxes, relations, HOIs, activities, and process steps.

Thanks to all great collaborators from @ramblr.ai

[Patrick Knab @p_knab]

@AmitLeViAI @icmlconf I feel you. We’ve also had strong scores several times, but then got rejected by the AC because one reviewer pushed against interpretability (performance tradeoff) and the AC followed that line.

[Patrick Knab @p_knab]

We structure the design space of CBMs, disentangle key components, discuss recurring challenges, and outline a roadmap for how the field could evolve.
If we missed any relevant work published up to the end of 2025, please let us know—we will incorporate it in the next version.

[Patrick Knab @p_knab]

What’s in the Bottle? A Survey and Roadmap of Concept Bottleneck Models

CBMs are a rapidly growing direction in interpretable-by-design machine learning. However, the field has become increasingly fragmented.

Preprint Link:
https://t.co/SHspTEW8P2

[Patrick Knab @p_knab]

Together with David Steinmann , Udo Schlegel, and @WolfStammer , we decided it was time to take a step back: we review and categorize more than 100 papers into a unified architectural taxonomy. In doing so, we try to answer the question: What’s in the bottle?

[Patrick Knab @p_knab]

@icmlconf As soon as ICLR decisions are out, amount will roughly decrease by number of accepted papers

[Patrick Knab @p_knab]

We wish everyone the best for tomorrow’s decisions—and try not to take them too seriously. 🙂

If you’re curious what tabular ML predicts for your paper, you can check it out here:
Website: https://t.co/xzS97oGNcB
Code: https://t.co/KleJBcpVUr

[Patrick Knab @p_knab]

We took a quick look at the new PaperDecision results for ICLR 2026 and asked a simple question: do we really need LLMs for this task?

We did the same but with tabular data:
https://t.co/rx1P8BU03G

#ICLR2026

[Patrick Knab @p_knab]

Framing the problem as primarily tabular—using numerical review scores such as soundness and contribution—we evaluated standard tabular models (TabPFN, CatBoost, logistic regression, and decision trees) without any hyperparameter tuning.

[Patrick Knab @p_knab]

@dyhTHU Super cool. We have built a similar tool, but based on tabular data. Let's see which of the two works best ;)
https://t.co/rx1P8BU03G

[Patrick Knab @p_knab]

@HeNordlinder @dyhTHU We build sth similar with more benchmarks and completely based on tabular data ;)
https://t.co/rx1P8BU03G

[Patrick Knab @p_knab]

@bfl_ml Congrats! Also a huge milestone for AI in south Germany :)
Just out of curiosity ( I haven’t been able to find sth on the website), are you also offering PhD internships?

[Patrick Knab @p_knab]

@rdesh26 @iclr_conf Sure, totally agree. We'll prepare a brief summary for the AC to summarize the work of our rebuttal, since we had some useful discussions.