👋 Hi, I'm Philipp!
I'm a NLP researcher fascinated by how we can teach LLMs to reason and tackle complex, structured problems. My research spans from multilingual NLP over information extraction to my current work on reasoning in AI for Math at Huawei in London. I completed my PhD at KU Leuven, where my research focused on multilinguality and NLP for Business applications.

What excites me

---

📄 Selected Publications & Projects

Conjecturing: An Overlooked Step in Formal Mathematical Reasoning

Jasivan Alex Sivakumar, Philipp Borchert, Ronald Cardenas, Gerasimos Lampouras

Preprint 2025

We identify conjecturing as a critical bottleneck in formal mathematical reasoning. We introduce ConjectureBench, a dataset to evaluate LLM conjecturing capabilities, and LEAN-FIRE, an inference-time method that improves autoformalisation by treating conjecturing as a distinct step.

📝 ArXiv

TopoAlign: A Framework for Aligning Code to Math via Topological Decomposition

Yupei Li, Philipp Borchert, Gerasimos Lampouras

Preprint 2025

TopoAlign addresses data scarcity for Math LLMs by aligning code to mathematical structures. It decomposes code repositories into components that mirror formal statements in Lean 4 for pretraining. The framework significantly improves performance on autoformalization benchmarks like MiniF2F and ProofNet.

📝 ArXiv

DRIFT: Decompose, Retrieve, Illustrate, then Formalize Theorems

Meiru Zhang, Philipp Borchert, Milan Gritta, Gerasimos Lampouras

Preprint 2025

DRIFT decomposes informal statements, retrieves dependent premises, and selects illustrative examples before formalization. This framework achieves state-of-the-art results on ProofNet and out-of-distribution benchmarks like ConNF.

📝 ArXiv

Language Fusion for Parameter-Efficient Cross-lingual Transfer

Philipp Borchert, Ivan Vulić, Marie-Francine Moens, Jochen De Weerdt

ACL 2025 | Vienna, Austria 📍

This study introduces Fusion for Language Representations (FLARE), a novel method that merges source and target language representations within low-rank adapters, enhancing cross-lingual transfer performance while maintaining parameter efficiency.

💻 GitHub 📝 ArXiv

Self-Distillation for Model Stacking Unlocks Cross-Lingual NLU in 200+ Languages

Fabian David Schmidt, Philipp Borchert, Ivan Vulić, Goran Glavaš

EMNLP (Findings) 2024 | Miami, USA 📍

This paper introduces MT-LLM, a method integrating machine translation (MT) encoders into LLM backbones using self-distillation. This approach unlocks natural language understanding capabilities for over 127 languages.

💻 GitHub 📝 ArXiv

Efficient Information Extraction in Few-Shot Relation Classification

Philipp Borchert, Jochen De Weerdt, Marie-Francine Moens

NAACL 2024 | Mexico City, Mexico 📍

This paper presents MultiRep, a novel approach to improve few-shot relation classification by combining multiple sentence representations using contrastive learning.

💻 GitHub 📝 ArXiv

CORE: A Few-Shot Company Relation Classification Dataset

Philipp Borchert, Jochen De Weerdt, Kristof Coussement, Arno De Caigny, Marie-Francine Moens

EMNLP 2023 | Singapore 📍

CORE is a dataset for few-shot relation classification focused on company relations, challenging models with high contextual complexity.

💻 GitHub 📝 ArXiv

Investigating Bias in Multilingual Language Models

Manon Reusens, Philipp Borchert, Margot Mieskes, Jochen De Weerdt, Bart Baesens

EMNLP 2023 | Singapore 📍

This study investigates whether debiasing techniques can be effectively transferred across different languages within multilingual LLMs.

💻 GitHub 📝 ArXiv

SEER: A Knapsack approach to Exemplar Selection

Jonathan Tonglet, Manon Reusens, Philipp Borchert, Bart Baesens

EMNLP 2023 | Singapore 📍

This paper introduces SEER, a novel method for selecting diverse and representative examples for in-context learning in complex question-answering tasks.

💻 GitHub 📝 ArXiv