Skip to content
View ben300694's full-sized avatar
🐟
🐟
11 followers · 35 following

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx3Achievement: YOLOAchievement: Arctic Code Vault Contributor

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx3Achievement: YOLOAchievement: Arctic Code Vault Contributor

Highlights

  • Pro

Block or report ben300694

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ben300694/README.md

Ben Ruppik

Postdoctoral Researcher at the intersection of Topological Deep Learning and Representation Learning — applying Topological Data Analysis (TDA) to Natural Language Processing and Task-Oriented Dialogue Systems.

  • 🏛️ Member of the Dialog Systems & Machine Learning Lab (Prof. Milica Gašić), Heinrich-Heine-Universität Düsseldorf
  • 🎓 Previously: PhD in Low-Dimensional Topology (Max Planck Institute for Mathematics Bonn & University of Bonn)
  • 🙂 Pronouns: he/him

🔬 Selected projects

Less is More: Local Intrinsic Dimensions of Contextual Language Models
Benjamin Ruppik et al., NeurIPS 2025
We analyze the geometry of contextual embedding spaces via local intrinsic dimension (LID) to track phenomena like overfitting and grokking; decreasing mean LID aligns with performance gains.
NeurIPS page · arXiv:2506.01034 · 📦 Code: Topo_LLM_public · grokking-via-lid

Local Topology Measures of Contextual Language Model Latent Spaces With Applications to Dialogue Term Extraction
Benjamin Ruppik et al., SIGDIAL 2024 — Best Paper Nomination
Contextual topological features derived from a corpus improve a tagging task on dialogue data.
ACL Anthology · doi:10.18653/v1/2024.sigdial-1.31 · arXiv:2408.03706 · 📦 Code: tda4contextualembeddings-public (GitLab)

🌐 Find me

Website · GitHub · HuggingFace · LinkedIn · Bluesky · Xing · X · Math.SE

📄 Papers

arXiv · ORCID · Google Scholar · ResearchGate · ACL Anthology

Pinned Loading

  1. aidos-lab/Topo_LLM_public aidos-lab/Topo_LLM_public Public

    Investigating embedding spaces generated by language models from a topological perspective via local intrinsic dimension (LID).

    Python 3

  2. aidos-lab/grokking-via-lid aidos-lab/grokking-via-lid Public

    Computing Local Intrinsic Dimensions (LID) to detect Grokking of Transformers.

    Python 1

  3. word-embeddings word-embeddings Public

    Repository for the seminar "Word Embedding Spaces", Master CS and Master AI & Data Science @ Heinrich Heine University Düsseldorf.

    HTML 3

  4. knot-theory knot-theory Public

    Code related to low-dimensional topology: Algorithms for Computing Invariants of Trisected Branched Covers and bounds for the Casson-Whitney Unknotting Number.

    Mathematica 5

  5. torsion-in-gamma torsion-in-gamma Public

    SageMath module to calculate the invariant Tors (Gamma pi_2 K) / pi_1 K for specific CW-complexes

    Python

  6. semanticLabelingTool semanticLabelingTool Public

    Forked from mgarbade/semanticLabelingTool

    Tool to create ground truth semantic segmentation masks using super pixels

    MATLAB