Language models
We study modern large language models as well as conventional statistical and rule-based language models: their architecture, domain adaptation, evaluation, and reliable integration into real systems.
Research at AI Værk
We do more than build with AI. We investigate how models and algorithmic systems actually work.
Our research combines technical depth with methods others can follow. We publish the question, data basis, results, and limitations so the work can be examined, challenged, and extended.
OpenMethods and assumptions visible
EmpiricalClaims grounded in data
AppliedResearch connected to practice
Fields
What we investigate.
AI models and evaluation
How can model quality be measured rigorously? We work on benchmarks, failure patterns, calibration, and methods that deliver more than good averages: they make decisions traceable.
Algorithmic systems
AI never acts in isolation. We also examine markets, platforms, and decision environments in which models, automation, and human behavior meet.
Publications
Methods before claims.
Publication 01 · Prediction markets
Who Trades on Polymarket?
A heuristic classification and comparative analysis of trader types on a prediction market.
The study classifies more than one million pseudonymous wallets from their observable trading behavior. It shows how sharply market weight, profit, and forecast quality diverge across Retail, Informed Traders, Whales, Algorithmic Traders, and Market Makers.
Read the study1,035,224classified wallets
261.6mtrades analyzed
$26.86bntrading volume
A research question with practical relevance?
We are open to conversations about joint experiments, datasets, evaluations, and research collaborations.