Stefan Feuerriegel

Generative orthogonal learning unlocks rate double robustness and quasi-oracle efficiency for estimating distributional treatment effects under complex time-varying confounding.

This work designs two interventions to reduce offloading decisions: metacognitive feedback that makes the implications of offloading for users explicit, and an effort-based reward that incentivizes less extensive LLM assistance.

Multilingual LLMs may misinterpret surface cues as genuine cultural understanding, risking the perpetuation of biases in AI outputs.

LLMs with web access only marginally outperform their offline counterparts, challenging assumptions about the necessity of real-time data for accurate predictions.

AutoSynthesis achieves expert-level meta-analysis with automated precision, making evidence synthesis scalable and accessible.

OncoSynth slashes treatment effect estimation errors by up to 66% in oncology, transforming how synthetic data can inform precision medicine.

LLMs can automate reproducibility assessments, achieving a 41% recovery rate of effect sizes—outperforming human analysts.

Men are 6 percentage points more likely to see ads from populist parties, raising alarms about gender bias in political ad delivery on social media.

LLM development is flying blind by ignoring causal inference, leaving models vulnerable to confounding and distribution shifts throughout pretraining, alignment, and evaluation.

Naive neural operator estimates of solution functional quantities can be significantly biased, but this paper provides a surprisingly simple debiasing technique to fix it.

Unstable estimates in long-term treatment effect prediction? These orthogonal learners use custom overlap weights to improve robustness in low-overlap settings common in long-term outcomes.