Institution profile

IT University of Copenhagen

Academic institutioneurope · dk
Official website
Research library265linked papers
Opportunities0open roles
Selected work

Representative Papers

What Are They Filtering Out? A Survey of Filtering Strategies for Harm Reduction in Pretraining Datasets

Feb 17, 2025arXiv.org

Pretraining data filtering strategies intended to reduce harmful content inadvertently exacerbate representational underrepresentation of marginalized groups, thereby amplifying demographic bias at the data level. Method: We systematically reviewed 55 English-language large language model technical reports to construct the first integrated data governance evaluation framework balancing safety and fairness. Through controlled experiments and quantitative bias analysis across mainstream filtering strategies, we measured their impact on group-level representation. Contribution/Results: Our analysis reveals that such strategies reduce text associated with disadvantaged groups by 12.7%–38.4% on average—significantly worsening representational disparity. This study provides the first empirical evidence refuting the “safety implies fairness” assumption in AI governance. We propose a co-optimization paradigm that jointly addresses content safety and equitable group representation, advocating for fairness-aware data curation in foundation model development.

2 citationsRead paper

Exact solutions to the Weighted Region Problem

Feb 19, 2024arXiv.org

This paper investigates the exact computability of shortest paths in weighted rectangular domains. In the rational algebraic computation model, we establish—for the first time—that the globally shortest path in a single rectangular domain with piecewise nonnegative weights (where path cost equals Euclidean length multiplied by weight) is algorithmically undecidable. Method: For source points located either on the boundary or in the interior, we explicitly construct and derive algebraic equations for bisectors in the shortest path map (SPM); their coefficients are rational functions of the input parameters. Leveraging algebraic computation theory, implicit curve analysis, and structural characterization of SPMs, we develop a complete analytic framework for exact shortest paths. Results: Our work rigorously delineates the boundary of exact solvability for this problem and provides the first bisector computation framework implementable within the rational algebraic model.

2 citationsRead paper

Improving Reasoning Performance in Large Language Models via Representation Engineering

Apr 28, 2025

This work addresses the challenge of precisely controlling large language models’ (LLMs) reasoning capabilities without fine-tuning. We propose a representation-level intervention method that identifies task-relevant activations within the residual stream, constructs task-specific control vectors from them, and directly modifies representations during inference to enhance inductive, deductive, and mathematical reasoning. To our knowledge, this is the first systematic application of representation engineering to LLM reasoning control—relying solely on forward-pass activation extraction and residual-stream intervention, thereby revealing the intrinsic decomposability of reasoning abilities. Evaluated on Mistral-7B-Instruct and Pythia models, our approach consistently improves accuracy across diverse reasoning benchmarks, stabilizes logit distributions, and its mechanistic validity is confirmed via KL divergence and entropy analysis. All code and analytical tools are publicly released to support reproducible representation intervention research.

1 citationsRead paper
Recent publications

Latest Papers