Publications related to Framing the News: From Human Perception to Large Language Model Inferences

Towards Motion Forecasting with Real-World Perception Inputs: Are End-to-End Approaches Competitive?

Motion forecasting is crucial in enabling autonomous vehicles to anticipate the future trajectories of surrounding agents. To do so, it requires solving mapping, detection, tracking, and then forecasting problems, in a multi-step pipeline. In this complex ...

2024

Social-Transmotion: Promptable Human Trajectory Prediction

Alexandre Massoud Alahi, Saeed Saadatnejad, Yang Gao, Kaouther Messaoud Ben Amor

Accurate human trajectory prediction is crucial for applications such as autonomous vehicles, robotics, and surveillance systems. Yet, existing models often fail to fully leverage the non-verbal social cues human subconsciously communicate when navigating ...

2024

Infusing structured knowledge priors in neural models for sample-efficient symbolic reasoning

Mattia Atzeni

The ability to reason, plan and solve highly abstract problems is a hallmark of human intelligence. Recent advancements in artificial intelligence, propelled by deep neural networks, have revolutionized disciplines like computer vision and natural language ...

EPFL2024

Serial Dependence in Human Visual Perception and Decision-Making

Gizay Ceylan

Decisions about a current visual stimulus are systematically biased by recently encountered stimuli, a phenomenon known as serial dependence. In human vision, for instance, we tend to report the features of current images as more similar â i.e., an attra ...

EPFL2024

Fly-Swat or Cannon? Cost-Effective Language Model Choice via Meta-Modeling

Maxime Jean Julien Peyrard, Robert West, Marija Sakota

Generative language models (LMs) have become omnipresent across data science. For a wide variety of tasks, inputs can be phrased as natural language prompts for an LM, from whose output the solution can then be extracted. LM performance has consistently be ...

Assoc Computing Machinery2024

Temporal aspects of feature integration in vision

Lukas Vogelsang

To fully comprehend visual perception, we need to necessarily understand its temporal dimension. Our visual environment is highly dynamic, requiring the processing and integration of temporal signals in order to make sense of it. Many processes, such as th ...

EPFL2024

Driving and suppressing the human language network using large language models

Martin Schrimpf

Transformer models such as GPT generate human-like language and are predictive of human brain responses to language. Here, using functional-MRI-measured brain responses to 1,000 diverse sentences, we first show that a GPT-based encoding model can predict t ...

Nature Portfolio2024

Probing and modulating inter-areal coupling in the cortical visual motion processing pathway with non-invasive brain stimulation

Michele Bevilacqua

In the last few years, stroke ranked as the second most common cause of death and is the third most significant condition affecting disability-adjusted life years (DALYs) worldwide. Being the most prevalent and quality of life impacting post-stroke symptom ...

EPFL2024

Understanding the sentiment associated with cultural ecosystem services using images and text from social media

Devis Tuia

Social media is increasingly being employed to develop Cultural Ecosystem Services (CES) indicators. The image-sharing platform Flickr has been one of the most popular sources of data. Most large-scale studies, however, tend to only use the number of image ...

2024

Revisiting Offline Compression: Going Beyond Factorization-based Methods for Transformer Language Models

Karl Aberer, Rémi Philippe Lebret, Mohammadreza Banaei

Recent transformer language models achieve outstanding results in many natural language processing (NLP) tasks. However, their enormous size often makes them impractical on memory-constrained devices, requiring practitioners to compress them to smaller net ...

Assoc Computational Linguistics-Acl2023

Modeling Structured Data in Attention-based Models

Alireza Mohammadshahi

Natural language processing has experienced significant improvements with the development of Transformer-based models, which employ self-attention mechanism and pre-training strategies. However, these models still present several obstacles. A notable issue ...

EPFL2023

Multi-task prompt-RSVQA to explicitly count objects on aerial images

Devis Tuia, Christel Marie Tartini-Chappuis, Nicola Antonio Santacroce, Sylvain Lobry, Javiera Francisca Castillo Navarro

Introduced to enable a wider use of Earth Observation images using natural language, Remote Sensing Visual Question Answering (RSVQA) remains a challenging task, in particular for questions related to counting. To address this specific challenge, we propos ...

2023

Text Representation Learning for Low Cost Natural Language Understanding

Jan Frederik Jonas Florian Mai

Natural language processing and other artificial intelligence fields have witnessed impressive progress over the past decade. Although some of this progress is due to algorithmic advances in deep learning, the majority has arguably been enabled by scaling ...

EPFL2023

XTab: Cross-table Pretraining for Tabular Transformers

Mahsa Shoaran, Bingzhao Zhu

The success of self-supervised learning in computer vision and natural language processing has motivated pretraining methods on tabular data. However, most existing tabular self-supervised learning models fail to leverage information across multiple data t ...

JMLR.org2023

Backpropagation-free training of deep physical neural networks

Romain Christophe Rémy Fleury, Babak Rahmani, Ali Momeni, Marc Philipp Del Hougne, Matthieu Francis Malléjac

Recent successes in deep learning for vision and natural language processing are attributed to larger models but come with energy consumption and scalability issues. Current training of digital deep-learning models primarily relies on backpropagation that ...

American Association for the Advancement of Science2023

Improving Generalization of Pretrained Language Models

Rabeeh Karimi Mahabadi

In this dissertation, we propose multiple methods to improve transfer learning for pretrained language models (PLMs). Broadly, transfer learning is a powerful technique in natural language processing, where a language model is first pre-trained on a data-r ...

EPFL2023

Deep Generative Models for Autonomous Driving: from Motion Forecasting to Realistic Image Synthesis

Saeed Saadatnejad

Forecasting is a capability inherent in humans when navigating. Humans routinely plan their paths, considering the potential future movements of those around them. Similarly, to achieve comparable sophistication and safety, autonomous systems must embrace ...

EPFL2023

The full-body illusion changes visual depth perception

Bruno Herbelin, Olaf Blanke, Sophie Jacqueline Andrée Betka

Knowing where objects are relative to us implies knowing where we are relative to the external world. Here, we investigated whether space perception can be influenced by an experimentally induced change in perceived self-location. To dissociate real and ap ...

Nature Research2023

TiC-SAT: Tightly-coupled Systolic Accelerator for Transformers

David Atienza Alonso, Giovanni Ansaloni, Alireza Amirshahi, Joshua Alexander Harrison Klein

Transformer models have achieved impressive results in various AI scenarios, ranging from vision to natural language processing. However, their computational complexity and their vast number of parameters hinder their implementations on resource-constraine ...

2023

Empathetic Conversational Agents for Distress Support

Kalpani Anuradha Welivita

Due to the increasing demands of today's fast-paced world, mental health concerns are on the rise, which necessitates innovative approaches to provide support to those in need. Open-domain conversational agents known as chatbots, offer a unique opportunit ...

EPFL2023