Author

Florian Mai

h-index 2 View on Semantic Scholar

We have 6 of 15 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.

#diffusion models Open access Sep 2026

Modeling of CO2-Based Fischer–Tropsch Synthesis over a Cu/Zn/K-Promoted Fe Catalyst: Influence of Reaction Kinetics and Multi-Fixed-Bed Reactor Design

CO2 hydrogenation by reverse water–gas shift (RWGS) directly combined with subsequent Fischer–Tropsch synthesis (FTS) in a single fixed-bed reactor represents a promising route for converting renewable hydrogen and captured carbon dioxide into hydrocarbons. However, the attainable CO2 conversion is limited by thermodyn...

Florian Mai, Andreas Jess · C++ Conference · 0 citations

#artificial intelligence Preprint May 2026

Detecting Hidden Behaviors in LLMs via Activation-matched Finetuning

Large language models can hide hidden behaviors that activate only under narrow conditions, such as backdoor triggers, sleeper-agent deployment cues, sandbagging, or topic-conditioned censorship. Such behaviors are difficult to detect without prior knowledge what to look for. We present activation-matched finetuning, a...

Robin Haselhorst, Lucie Flek, Florian Mai · 0 citations

#natural language process... Preprint May 2026

Reinforcement Learning Can Amplify Emergent Misalignment from Harmless Rewards

It is shown that rewarding narrow, overtly misaligned behavior produces substantially higher general-domain misalignment than sample-matched SFT, and that EM from RL can be induced by reward signals that could plausibly arise naturally, such as unpopular aesthetic preferences or poor rhetorical appeals.

Magnus Jørgenvåg, David Kaczér, Lasse Ruttert et al. · 2 citations

Open access Jul 2026

All too perfect: bias and aspiration in persona generation with LLMs

It is proposed that persona-based evaluation can serve as a scalable diagnostic of what generative systems value and prioritize when depicting humanity, and that persona generations are far from neutral.

N. Corrêa, Rafaela Weber Mallmann, David Kaczér et al. · Artificial Intelligence Revi... · 0 citations

Preprint Jul 2026

A Unified Moral-Value Dataset for Instruction Tuning

This work constructs a unified moral-value dataset that can be directly used for instruction tuning, built upon existing moral-value datasets by merging them into a unified corpus and converting them into an instruction-response format.

Zhaohui Zeng, Florian Mai · 0 citations

#natural language process... Preprint Aug 2026

Data Attribution of Emergent Misalignment with Persona Features

Emergent misalignment (EM) is the phenomenon where fine-tuning a language model on a narrow task leads to harmful behavior in unrelated domains. A leading mechanistic account attributes EM to persona features: latent directions acquired during pre-training that misaligned fine-tuning amplifies. We ask where these featu...

C. Vetter, David Kaczér, Lucie Flek et al. · 0 citations