Zeynep Akata

MemLoRA: Distilling Expert Adapters for On-Device Memory Systems

Massimo Bini, Ondrej Bohdal, Umberto Michieli +3

Memory-augmented Large Language Models (LLMs) have demonstrated remarkable consistency during prolonged dialogues by storing relevant memories and incorporating them as context. Su…

A Systematic Study of In-the-Wild Model Merging for Large Language Models

Oğuz Kağan Hitit, Leander Girrbach, Zeynep Akata

Model merging combines multiple fine-tuned checkpoints into a single model without additional training, offering an attractive approach to reusing models and efficiently improving…

Sparse Autoencoders are Topic Models

Leander Girrbach, Zeynep Akata

Sparse autoencoders (SAEs) are used to analyze embeddings, but their role and practical value are debated. We propose a new perspective on SAEs by demonstrating that they can be na…

Person-Centric Annotations of LAION-400M: Auditing Bias and Its Transfer to Models

Leander Girrbach, Stephan Alaniz, Genevieve Smith +2

Vision-language models trained on large-scale multimodal datasets show strong demographic biases, but the role of training data in producing these biases remains unclear. A major b…

Training-free Uncertainty Guidance for Complex Visual Tasks with MLLMs

Sanghwan Kim, Rui Xiao, Stephan Alaniz +2

Multimodal Large Language Models (MLLMs) often struggle with fine-grained perception, such as identifying small objects in high-resolution images or detecting key moments in long v…

Stitch: Training-Free Position Control in Multimodal Diffusion Transformers

Jessica Bader, Mateusz Pach, Maria A. Bravo +2

Text-to-Image (T2I) generation models have advanced rapidly in recent years, but accurately capturing spatial relationships like "above" or "to the right of" poses a persistent cha…

Reference-Free Rating of LLM Responses via Latent Information

Leander Girrbach, Chi-Ping Su, Tankred Saanum +3

How reliable are single-response LLM-as-a-judge ratings without references, and can we obtain fine-grained, deterministic scores in this setting? We study the common practice of as…

Road Obstacle Video Segmentation

Shyam Nandan Rai, Shyamgopal Karthik, Mariana-Iuliana Georgescu +3

With the growing deployment of autonomous driving agents, the detection and segmentation of road obstacles have become critical to ensure safe autonomous navigation. However, exist…

Noise Hypernetworks: Amortizing Test-Time Compute in Diffusion Models

Luca Eyring, Shyamgopal Karthik, Alexey Dosovitskiy +2

The new paradigm of test-time scaling has yielded remarkable breakthroughs in Large Language Models (LLMs) (e.g. reasoning models) and in generative vision models, allowing models…

SUB: Benchmarking CBM Generalization via Synthetic Attribute Substitutions

Jessica Bader, Leander Girrbach, Stephan Alaniz +1

Concept Bottleneck Models (CBMs) and other concept-based interpretable models show great promise for making AI applications more transparent, which is essential in fields like medi…

Investigating Structural Pruning and Recovery Techniques for Compressing Multimodal Large Language Models: An Empirical Study

Yiran Huang, Lukas Thede, Massimiliano Mancini +2

While Multimodal Large Language Models (MLLMs) demonstrate impressive capabilities, their substantial computational and memory requirements pose significant barriers to practical d…

Unlocking Pretrained Vision Transformers for Time Series Classification

Simon Roschmann, Quentin Bouniot, Vasilii Feofanov +2

Adapting vision models for time series analysis is compelling, yet all existing approaches are falling short of dedicated time series foundation models (TSFMs) in classification. I…

Align-then-Unlearn: Embedding Alignment for LLM Unlearning

Philipp Spohn, Leander Girrbach, Jessica Bader +1

As large language models (LLMs) are trained on massive datasets, they have raised significant privacy and ethical concerns due to their potential to inadvertently retain sensitive…

Subspace-Boosted Model Merging

Ronald Skorobogat, Karsten Roth, Mariana-Iuliana Georgescu

Model merging enables the combination of multiple specialized expert models into a single model capable of performing multiple tasks. However, the benefits of merging an increasing…

LoFT: LoRA-fused Training Dataset Generation with Few-shot Guidance

Jae Myung Kim, Stephan Alaniz, Cordelia Schmid +1

Despite recent advances in text-to-image generation, using synthetically generated data seldom brings a significant boost in performance for supervised learning. Oftentimes, synthe…

cs.AI2025

Feasibility with Language Models for Open-World Compositional Zero-Shot Learning

Jae Myung Kim, Stephan Alaniz, Cordelia Schmid +1

Humans can easily tell if an attribute (also called state) is realistic, i.e., feasible, for an object, e.g. fire can be hot, but it cannot be wet. In Open-World Compositional Zero…

DD-Ranking: Rethinking the Evaluation of Dataset Distillation

Zekai Li, Xinhao Zhong, Samir Khaki +49

In recent years, dataset distillation has provided a reliable solution for data compression, where models trained on the resulting smaller synthetic datasets achieve performance co…

Concept-Guided Interpretability via Neural Chunking

Shuchen Wu, Stephan Alaniz, Shyamgopal Karthik +3

Neural networks are often described as black boxes, reflecting the significant challenge of understanding their internal workings and interactions. We propose a different perspecti…

Sparse Autoencoders Learn Monosemantic Features in Vision-Language Models

Mateusz Pach, Shyamgopal Karthik, Quentin Bouniot +2

Sparse Autoencoders (SAEs) have recently gained attention as a means to improve the interpretability and steerability of Large Language Models (LLMs), both of which are essential f…

A Large Scale Analysis of Gender Biases in Text-to-Image Generative Models

Leander Girrbach, Stephan Alaniz, Genevieve Smith +1

With the increasing use of image generation technology, understanding its social biases, including gender bias, is essential. This paper presents a large-scale study on gender bias…

DeLoRA: Decoupling Angles and Strength in Low-rank Adaptation

Massimo Bini, Leander Girrbach, Zeynep Akata

Parameter-Efficient FineTuning (PEFT) methods have recently gained significant popularity thanks to the widespread availability of large-scale pretrained models. These methods allo…

WikiBigEdit: Understanding the Limits of Lifelong Knowledge Editing in LLMs

Lukas Thede, Karsten Roth, Matthias Bethge +2

Keeping large language models factually up-to-date is crucial for deployment, yet costly retraining remains a challenge. Knowledge editing offers a promising alternative, but metho…