Author

Wojciech M. Czarnecki

h-index 35 View on Semantic Scholar

We have 1 of 78 papers

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Preprint Aug 2026

Why Large Language Models Fail at Tabular Prediction

The results show that the LLM's capability dissolves with dimension in a way no noise-corrupted classical learner mimics - which explains why LLMs, so capable elsewhere, keep losing to fifty-year-old baselines on tables, while leaving the mechanism of the prediction as an open question.

M. Garnelo, Wojciech M. Czarnecki · 1 citation