Page Stream Segmentation with LLMs: Challenges and Applications in Insurance Document Automation
Explores LLM applications for document processing automation in insurance, using parameter-efficient fine-tuning for …
Research
My research focuses on natural language processing, AI safety, and machine learning applications in science and industry. I work on making AI systems more robust and useful across domains like document processing and computational science.
Featured Publications (3)
Deconstructing recurrence, attention, and gating: Investigating the transferability of transformers and gated recurrent neural networks in forecasting of dynamical systems
Ablation study examining key architectural components of neural networks for spatiotemporal forecasting, finding that …
The earth is flat and the sun is not a star: The susceptibility of GPT-2 to universal adversarial triggers
An investigation into whether universal adversarial triggers can control not just the topic but also the stance of …
Conference Publications (3)
Page Stream Segmentation with LLMs: Challenges and Applications in Insurance Document Automation
Explores LLM applications for document processing automation in insurance, using parameter-efficient fine-tuning for …
The earth is flat and the sun is not a star: The susceptibility of GPT-2 to universal adversarial triggers
An investigation into whether universal adversarial triggers can control not just the topic but also the stance of …
Investigating Coordinated 'Social' Targeting of High-Profile Twitter Accounts
An investigation into strange follower dynamics on high-profile Twitter accounts, documenting phenomena like sub-second …
Workshop Publications (1)
Latent semantic network induction in the context of linked example senses
Explores a data-driven approach to construct a WordNet-like semantic network using the entirety of the noisy, …
Preprints & Working Papers (6)
Deconstructing recurrence, attention, and gating: Investigating the transferability of transformers and gated recurrent neural networks in forecasting of dynamical systems
Ablation study examining key architectural components of neural networks for spatiotemporal forecasting, finding that …
Large Language Models for Page Stream Segmentation
We enhance the TABME benchmark for page stream segmentation, creating TABME++, and show that fine-tuned, decoder-based …
EigenNoise: A Contrastive Prior to Warm-Start Representations
A preliminary investigation into EigenNoise, a simple, data-free initialization scheme for word vectors that can …
To Know by the Company Words Keep and What Else Lies in the Vicinity
Presents an analytical model of statistics learned by Word2Vec and GloVe. This work derives the first known analytical …
NewsTweet: a dataset of social media embedding in online journalism
Describes the creation of NewsTweet, a large-scale dataset and pipeline for studying embedded tweets in online news. …
Investigating Coordinated 'Social' Targeting of High-Profile Twitter Accounts
An investigation into strange follower dynamics on high-profile Twitter accounts, documenting phenomena like sub-second …
Statistics
6
Preprints
3
Published Papers
9
Total Publications
5
Publication Venues