# Paulius Rauba

> Paulius Rauba — machine learning researcher and engineer. Founding Research Scientist at Cursive, recently PhD’d at Cambridge. Research, writing, and side projects across deep learning, generative models, and AI4Science.

Personal research site of Paulius Rauba — machine learning researcher, Founding Research Scientist at Cursive, recently PhD'd in ML at the University of Cambridge. Research areas: deep hierarchical learning, self-healing machine learning, context-aware testing of LLMs, statistical robustness audits for language models, generative AI4Science.

## Notes

- [All notes](https://www.pauliusrauba.com/notes): Published research and engineering notes.
- [RSS feed](https://www.pauliusrauba.com/notes/feed.xml): Full published notes and updates.
- [#1 Principles for setting up a cheap proxy model for reinforcement learning training](https://www.pauliusrauba.com/notes/1-principles-for-setting-up-a-cheap-proxy-model-for-reinforcement-learning-training): How to choose and validate a cheap proxy model for reinforcement learning: configurations, LoRA, policy lag, and experiments with Kimi K3.1 and Qwen.
  - [Markdown](https://www.pauliusrauba.com/notes/1-principles-for-setting-up-a-cheap-proxy-model-for-reinforcement-learning-training.md). Published 2026-11-13T14:54:23.361Z; updated 2026-11-13T15:15:48.519Z.

- [Full text index](https://www.pauliusrauba.com/llms-full.txt): Published notes and research summaries.

## Interactive paper expositions

- [Nested Subspace Networks (ICLR 2026)](https://www.pauliusrauba.com/papers/nested-subspace-networks): Nested Subspace Networks (NSNs) reparameterize each linear layer of a neural network as a low-rank product W = B A whose columns are nested, so that every rank-r sub-model is a strict prefix of every higher-rank one.
- [Self-Healing Machine Learning (NeurIPS 2024)](https://www.pauliusrauba.com/papers/self-healing-ml): Self-Healing Machine Learning (SHML) is a framework for models that autonomously adapt to distribution shift without human intervention.
- [Context-Aware Testing (NeurIPS 2024)](https://www.pauliusrauba.com/papers/context-aware-testing): Context-Aware Testing (CAT) is a new paradigm for ML model testing that uses large language models to read the surrounding context of a task and generate the failure modes that actually matter, rather than exhaustively scanning data slices.
- [Auditing LLM Robustness (DBPA) (ICML 2025)](https://www.pauliusrauba.com/papers/auditing-llms): Distribution-Based Perturbation Analysis (DBPA) is a frequentist hypothesis-testing framework for auditing the robustness of large language models.

## Canonical paper records

- [Deep Hierarchical Learning with Nested Subspace Networks for Large Language Models — Paulius Rauba, Mihaela van der Schaar, ICLR 2026](https://openreview.net/forum?id=ymUOPsbxLi)
- [Self-Healing Machine Learning: A Framework for Autonomous Adaptation in Real-World Environments — Paulius Rauba, Nabeel Seedat, Krzysztof Kacprzyk, Mihaela van der Schaar, NeurIPS 2024](https://openreview.net/forum?id=f63DKIpx0I)
- [Context-Aware Testing: A New Paradigm for Model Testing with Large Language Models — Paulius Rauba, Nabeel Seedat, Max Ruiz Luyten, Mihaela van der Schaar, NeurIPS 2024](https://openreview.net/forum?id=d75qCZb7TX)
- [Statistical Hypothesis Testing for Auditing Robustness in Language Models — Paulius Rauba, Qiyao Wei, Mihaela van der Schaar, ICML 2025](https://openreview.net/forum?id=ECayXPDoha)

## Profiles

- https://www.linkedin.com/in/paulius-rauba/
- https://scholar.google.com/citations?user=NoVFmOgAAAAJ&hl=en
- https://github.com/pauliusrauba
- https://pauliusrauba.substack.com

## Contact

- Email: hello@pauliusrauba.com
- Substack: https://pauliusrauba.substack.com