Welcome to our new website

We’ve launched our new team website. You can find here our research directions, academic highlights, blog posts, and team info.

Academic Generalization & Interpretability

Not All Splits Are Equal: Rethinking Attribute Generalization Across Unrelated Categories

Liviu Nicolae Fircă, Antonio Bărbălau, Dan Oneata, Elena Burceanu

Accepted at NeurIPS 2025 Workshop CauScien (Poster) Dec 2025

Links: arXiv GitHub Abstract Can models generalize attribute knowledge across semantically and perceptually dissimilar categories? While prior work …

Academic Natural Language Processing

Rethinking Sparse Autoencoders: Select-and-Project for Fairness and Control from Encoder Features Alone

Antonio Bărbălau, Cristian Daniel Păduraru, Teodor Poncu, Alexandru Tifrea, Elena Burceanu

Accepted at NeurIPS 2025 Workshops: Mechanistic Interpretability; Reliable ML from Unreliable Data (Poster) Dec 2025

Links: arXiv Abstract Sparse Autoencoders (SAEs) have proven valuable due to their ability to provide interpretable and steerable representations. …

Academic Generalization & Interpretability

Learning (Approximately) Equivariant Networks via Constrained Optimization

Andrei Manolache, Luiz F.O. Chamon, Mathias Niepert

Accepted at NeurIPS 2025 (Oral, top 0.4%) Dec 2025

Links: arXiv · GitHub Abstract Equivariant neural networks are designed to respect symmetries through their architecture, boosting generalization and …

Patent

Systems and Methods of Detecting Bias in AI Training Data

rfilipescu, ext-abarbalau, eburceanu

Nov 2025

Docket: BTD-2414-C1 · US application: 19/378,395 · PTO filed: Nov 4, 2025 · Status: Filed · Technologies: Machine learning Continuation patent …

Patent

Systems and Methods of Identifying Spurious Correlations in AI Training Data

rfilipescu, ext-abarbalau, eburceanu

Nov 2025

Docket: BTD-2414 · US application: 19/378,391 · PTO filed: Nov 4, 2025 · Status: Filed · Technologies: Machine learning Patent application on …

Patent

Systems and Methods of Training Deepfake Detectors

eoneata, ssmeu

Sep 2025

Docket: BTD-2511 · PTO filed: Sep 30, 2025 · Status: Filed · Technologies: Machine learning US patent application on training deepfake detectors.

Academic Natural Language Processing

Learning the Neighborhood: Contrast-Free Multimodal Self-Supervised Molecular Graph Pretraining

Boshra Ariguib, Mathias Niepert, Andrei Manolache

Under review Sep 2025

Links: arXiv GitHub Abstract High-quality molecular representations are essential for property prediction and molecular design, yet large labeled …

News

NeurIPS Associate Program Chair

Sep 2025

A team member served as NeurIPS Associate Program Chair (Associate PC), helping shape the conference’s review process and paper selection.

News

CVPR Daily interview

Jun 2025

Our group was featured in a CVPR Daily interview during CVPR 2025. The interview can be found here

Academic Deepfake Detection

Circumventing shortcuts in audio-visual deepfake detection datasets with unsupervised learning

Stefan Smeu, Dragos-Alexandru Boldisor, Dan Oneata, Elisabeta Oneata

Accepted at CVPR 2025 (Highlight, top 3%) Jun 2025

Links: arXiv · GitHub Abstract Good datasets are essential for developing and benchmarking any machine learning system. Their importance is even more …