papers

Publications (40)

cs.CV2024

Reading Between the Frames: Multi-Modal Depression Detection in Videos from Non-Verbal Cues

David Gimeno-Gómez, Ana-Maria Bucur, Adrian Cosma +2

Depression, a prominent contributor to global disability, affects a substantial portion of the population. Efforts to detect depression from social media texts have been prevalent,…

cs.CV2024

Aligning Actions and Walking to LLM-Generated Textual Descriptions

Radu Chivereanu, Adrian Cosma, Andy Catruna +2

Large Language Models (LLMs) have demonstrated remarkable capabilities in various domains, including data augmentation and synthetic data generation. This work explores the use of…

cs.CV2020

Self-Supervised Representation Learning on Document Images

Adrian Cosma, Mihai Ghidoveanu, Michael Panaitescu-Liess +1

This work analyses the impact of self-supervised pre-training on document images in the context of document image classification. While previous approaches explore the effect of se…

cs.CV2024

Gait Recognition from Highly Compressed Videos

Andrei Niculae, Andy Catruna, Adrian Cosma +2

Surveillance footage represents a valuable resource and opportunities for conducting gait analysis. However, the typical low quality and high noise levels in such footage can sever…

cs.LG2020

A Generic and Model-Agnostic Exemplar Synthetization Framework for Explainable AI

Antonio Barbalau, Adrian Cosma, Radu Tudor Ionescu +1

With the growing complexity of deep learning methods adopted in practical applications, there is an increasing and stringent need to explain and interpret the decisions of such met…

cs.CV2026

Spatial Colour Mixing Illusions as a Perception Stress Test for Vision-Language Models

Nicoleta-Nina Basoc, Adrian Cosma, Emilian Radoi

Vision-language models (VLMs) achieve strong benchmark results, yet can exhibit systematic perceptual weaknesses: structured, large changes to pixel values can cause confident yet…

cs.CL2023

It's Just a Matter of Time: Detecting Depression with Time-Enriched Multimodal Transformers

Ana-Maria Bucur, Adrian Cosma, Paolo Rosso +1

Depression detection from user-generated content on the internet has been a long-lasting topic of interest in the research community, providing valuable screening tools for psychol…

cs.CV2025

MoME: Estimating Psychological Traits from Gait with Multi-Stage Mixture of Movement Experts

Andy Cǎtrunǎ, Adrian Cosma, Emilian Rǎdoi

Gait encodes rich biometric and behavioural information, yet leveraging the manner of walking to infer psychological traits remains a challenging and underexplored problem. We intr…

cs.CV2024

CrossGaze: A Strong Method for 3D Gaze Estimation in the Wild

Andy Cătrună, Adrian Cosma, Emilian Rădoi

Gaze estimation, the task of predicting where an individual is looking, is a critical task with direct applications in areas such as human-computer interaction and virtual reality.…

cs.LG2026

Humanity's Last Exam

Long Phan, Alice Gatti, Ziwen Han +1144

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achi…

cs.LG2026

Counterfactual Methods for Detecting Unfairness in Anti-Money Laundering Algorithms

Lea Multerer, Michele Inchingolo, David Kletz +3

The application of machine learning-based predictive algorithms to Anti-Money Laundering (AML) has grown rapidly, driven by the vast volume of financial transaction data available…

cs.CL2026

Automatic Prompt Optimization for Dataset-Level Feature Discovery

Adrian Cosma, Oleg Szehr, David Kletz +2

Feature extraction from unstructured text is a critical step in many downstream classification pipelines, yet current approaches largely rely on hand-crafted prompts or fixed featu…

cs.CL2025

The Strawberry Problem: Emergence of Character-level Understanding in Tokenized Language Models

Adrian Cosma, Stefan Ruseti, Emilian Radoi +1

Despite their remarkable progress across diverse domains, Large Language Models (LLMs) consistently fail at simple character-level tasks, such as counting letters in words, due to…

cs.CV2024

GaitPT: Skeletons Are All You Need For Gait Recognition

Andy Catruna, Adrian Cosma, Emilian Radoi

The analysis of patterns of walking is an important area of research that has numerous applications in security, healthcare, sports and human-computer interaction. Lately, walking…

cs.CV2023

Learning to Simplify Spatial-Temporal Graphs in Gait Analysis

Adrian Cosma, Emilian Radoi

Gait analysis leverages unique walking patterns for person identification and assessment across multiple domains. Among the methods used for gait analysis, skeleton-based approache…

cs.CL2025

Dr.Copilot: A Multi-Agent Prompt Optimized Assistant for Improving Patient-Doctor Communication in Romanian

Andrei Niculae, Adrian Cosma, Cosmin Dumitrache +1

Text-based telemedicine has become increasingly common, yet the quality of medical advice in doctor-patient interactions is often judged more on how advice is communicated rather t…

cs.CL2026

An In-Vitro Study on Cross-Lingual Generalization in Language Models

Adrian Cosma

Cross-lingual transfer in language models is difficult to study in natural corpora because lexical overlap, morphology, data imbalance, and tokenization are entangled. We introduce…

cs.CL2021

Sequence-to-Sequence Lexical Normalization with Multilingual Transformers

Ana-Maria Bucur, Adrian Cosma, Liviu P. Dinu

Current benchmark tasks for natural language processing contain text that is qualitatively different from the text used in informal day to day digital communication. This discrepan…

cs.CL2025

A Retrieval-Based Approach to Medical Procedure Matching in Romanian

Andrei Niculae, Adrian Cosma, Emilian Radoi

Accurately mapping medical procedure names from healthcare providers to standardized terminology used by insurance companies is a crucial yet complex task. Inconsistencies in namin…

cs.CL2022

Life is not Always Depressing: Exploring the Happy Moments of People Diagnosed with Depression

Ana-Maria Bucur, Adrian Cosma, Liviu P. Dinu

In this work, we explore the relationship between depression and manifestations of happiness in social media. While the majority of works surrounding depression focus on symptoms,…

cs.CL2022

BLUE at Memotion 2.0 2022: You have my Image, my Text and my Transformer

Ana-Maria Bucur, Adrian Cosma, Ioan-Bogdan Iordache

Memes are prevalent on the internet and continue to grow and evolve alongside our culture. An automatic understanding of memes propagating on the internet can shed light on the gen…

cs.CV2023

PsyMo: A Dataset for Estimating Self-Reported Psychological Traits from Gait

Adrian Cosma, Emilian Radoi

Psychological trait estimation from external factors such as movement and appearance is a challenging and long-standing problem in psychology, and is principally based on the psych…

cs.CV2021

From Face to Gait: Weakly-Supervised Learning of Gender Information from Walking Patterns

Andy Catruna, Adrian Cosma, Ion Emilian Radoi

Obtaining demographics information from video is valuable for a range of real-world applications. While approaches that leverage facial features for gender inference are very succe…

cs.CV2020

Black-Box Ripper: Copying black-box models using generative evolutionary algorithms

Antonio Barbalau, Adrian Cosma, Radu Tudor Ionescu +1

We study the task of replicating the functionality of black-box neural models, for which we only know the output class probabilities provided for a set of input images. We assume b…

cs.CL2026

Training Language Models with homotokens Leads to Delayed Overfitting

Adrian Cosma, Stefan Ruseti, Emilian Radoi +1

Subword tokenization introduces a computational layer in language models where many distinct token sequences decode to the same surface form and preserve meaning, yet induce differ…

cs.CL2024

RoCode: A Dataset for Measuring Code Intelligence from Problem Definitions in Romanian

Adrian Cosma, Bogdan Iordache, Paolo Rosso

Recently, large language models (LLMs) have become increasingly powerful and have become capable of solving a plethora of tasks through proper instructions in natural language. How…

cs.CV2024

The Paradox of Motion: Evidence for Spurious Correlations in Skeleton-based Gait Recognition Models

Andy Cătrună, Adrian Cosma, Emilian Rădoi

Gait, an unobtrusive biometric, is valued for its capability to identify individuals at a distance, across external outfits and environmental conditions. This study challenges the…

cs.CV2023

GaitFormer: Learning Gait Representations with Noisy Multi-Task Learning

Adrian Cosma, Emilian Radoi

Gait analysis is proven to be a reliable way to perform person identification without relying on subject cooperation. Walking is a biometric that does not significantly change in s…

cs.CV2025

Database-Agnostic Gait Enrollment using SetTransformers

Nicoleta Basoc, Adrian Cosma, Andy Cǎtrunǎ +1

Gait recognition has emerged as a powerful tool for unobtrusive and long-range identity analysis, with growing relevance in surveillance and monitoring applications. Although recen…

cs.CV2025

On Model and Data Scaling for Skeleton-based Self-Supervised Gait Recognition

Adrian Cosma, Andy Cǎtrunǎ, Emilian Rǎdoi

Gait recognition from video streams is a challenging problem in computer vision biometrics due to the subtle differences between gaits and numerous confounding factors. Recent adva…

cs.CV2021

WildGait: Learning Gait Representations from Raw Surveillance Streams

Adrian Cosma, Emilian Radoi

The use of gait for person identification has important advantages such as being non-invasive, unobtrusive, not requiring cooperation and being less likely to be obscured compared…

cs.CL2026

Improving Medical Communication using Rubric-Guided Counterfactual Recommendations

Adrian Cosma, Nicoleta-Nina Basoc, Andrei Niculae +2

Text-based telemedicine increasingly relies on lightweight patient feedback, however, such feedback primarily reflects perceived communication quality rather than medical accuracy.…

cs.CL2022

An End-to-End Set Transformer for User-Level Classification of Depression and Gambling Disorder

Ana-Maria Bucur, Adrian Cosma, Liviu P. Dinu +1

This work proposes a transformer architecture for user-level classification of gambling addiction and depression that is trainable end-to-end. As opposed to other methods that oper…

cs.CV2018

CamLoc: Pedestrian Location Detection from Pose Estimation on Resource-constrained Smart-cameras

Adrian Cosma, Ion Emilian Radoi, Valentin Radu

Recent advancements in energy-efficient hardware technology is driving the exponential growth we are experiencing in the Internet of Things (IoT) space, with more pervasive computa…

cs.CL2025

RoMath: A Mathematical Reasoning Benchmark in Romanian

Adrian Cosma, Ana-Maria Bucur, Emilian Radoi

Mathematics has long been conveyed through natural language, primarily for human understanding. With the rise of mechanized mathematics and proof assistants, there is a growing nee…

cs.CL2024

How Hard is this Test Set? NLI Characterization by Exploiting Training Dynamics

Adrian Cosma, Stefan Ruseti, Mihai Dascalu +1

Natural Language Inference (NLI) evaluation is crucial for assessing language understanding models; however, popular datasets suffer from systematic spurious correlations that arti…

cs.CL2026

Prompting Complexity: Shortest Prompts for Texts and Behaviors in LLMs

Adrian Cosma

In this paper, we define the quantity of prompting complexity: for a fixed instruction-tuned language model, what is the shortest plausible prompt that makes deterministic decoding…

cs.CV2023

GaitMorph: Transforming Gait by Optimally Transporting Discrete Codes

Adrian Cosma, Emilian Radoi

Gait, the manner of walking, has been proven to be a reliable biometric with uses in surveillance, marketing and security. A promising new direction for the field is training gait…

cs.CL2021

Early Risk Detection of Pathological Gambling, Self-Harm and Depression Using BERT

Ana-Maria Bucur, Adrian Cosma, Liviu P. Dinu

Early risk detection of mental illnesses has a massive positive impact upon the well-being of people. The eRisk workshop has been at the forefront of enabling interdisciplinary res…

cs.CL2026

What Makes a Good Doctor Response? A Study on Text-Based Telemedicine

Adrian Cosma, Cosmin Dumitrache, Emilian Radoi

Text-based telemedicine has become an increasingly used mode of care, requiring clinicians to deliver medical advice clearly and effectively in writing. As platforms increasingly r…