Papers by Timothee Mickus
MAMMOTH: Massively Multilingual Modular Open Translation @ Helsinki (2024.eacl-demo)
Copied to clipboard
Timothee Mickus, Stig-Arne Grönroos, Joseph Attieh, Michele Boggia, Ona De Gibert, Shaoxiong Ji, Niki Andreas Loppi, Alessandro Raganato, Raúl Vázquez, Jörg Tiedemann
| Challenge: | a growing trend towards modularization is limiting the size and information that can be handled in large language models. |
| Approach: | They propose a framework for training massively multilingual modular machine translation systems at scale. |
| Outcome: | The proposed framework is adapted to train multilingual models at scale on NVIDIA GPUs. |
Pre-trained Language Models Learn Remarkably Accurate Representations of Numbers (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing work showed limited success in probing numeric values from models’ representations, indicating that these errors can be attributed to the inherent unreliability of distributionally learned embeddings in representing exact quantities. |
| Approach: | They propose a probing technique that decodes numeric values from input embeddings with near-perfect accuracy across a range of open-source LMs. |
| Outcome: | The proposed probing technique decodes numeric values from input embeddings with near-perfect accuracy across a range of open-source LMs. |
The Emergence of High-Level Semantics in a Signaling Game (2024.starsem-1)
Copied to clipboard
| Challenge: | a symbol grounding problem has been raised in recent years in AI . we show that neural agents can communicate high-level semantic concepts . |
| Approach: | They propose to use an adversarial agent to train neural agents in a signaling game . they show that the agents can communicate high-level semantic concepts rather than low-level features . |
| Outcome: | The proposed method can learn to communicate high-level semantic concepts . it also produces an appropriate training signal when no other method is available . |
„Mann“ is to “Donna” as「国王」is to « Reine » Adapting the Analogy Task for Multilingual and Contextual Embeddings (2023.starsem-1)
Copied to clipboard
| Challenge: | a lack of comparable multilingual benchmarks and a consensual evaluation protocol for contextual models remains an open question. |
| Approach: | They propose a multilingual analogy dataset and evaluate human and contextual embedding performance. |
| Outcome: | The proposed dataset evaluates human and contextual embedding models on the analogy task. |
So many design choices: Improving and interpreting neural agent communication in signaling games (2023.findings-acl)
Copied to clipboard
| Challenge: | Emergent language games are experimental protocols designed to model how communication may arise among a group of agents. |
| Approach: | They propose to adopt a signaling game in which a sender is exposed to an image and generates a sequence of symbols that is transmitted to a receiver. |
| Outcome: | The proposed language improves when the sender is exposed to an image and generates a sequence of symbols that is transmitted to a receiver. |
A Comparison of Language Modeling and Translation as Multilingual Pretraining Objectives (2024.emnlp-main)
Copied to clipboard
| Challenge: | Pretrained language models (PLMs) display impressive performances and have captured the attention of the NLP community. |
| Approach: | They propose to compare multilingual pretraining objectives in a controlled methodological environment with multilingual models. |
| Outcome: | The proposed model outperforms existing models in 6 languages and demonstrates that multilingual translation is an effective pretraining objective under the right conditions. |
Language Models Learn Universal Representations of Numbers and Here’s Why You Should Care (2026.acl-long)
Copied to clipboard
Michal Štefánik, Timothee Mickus, Marek Kadlčík, Bertram Højer, Michal Spiegel, Raúl Vázquez, Aman Sinha, Josef Kuchař, Philipp Mondorf, Pontus Stenetorp
| Challenge: | Prior work has shown that large language models (LLMs) often converge to accurate input embedding for numbers, based on sinusoidal representations. |
| Approach: | They show that large language models often converge to accurate input embedding for numbers, based on sinusoidal representations. |
| Outcome: | The proposed representations are strikingly systematic, and are interchangeable in a large swathe of experimental setups. |
Grounded and well-rounded: a methodological approach to the study of cross-modal and cross-lingual grounding (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies on grounding have focused on qualitatively different generalizations, but limited empirical evidence supports either position. |
| Approach: | They propose a methodological framework for studying the effects of grounding on NLP systems . they use a sample of models trained on different input modalities to tease out qualitative differences . |
| Outcome: | The proposed framework teases out qualitative differences in model behavior between models trained on different input sources from quantifiable models. |
Isotropy, Clusters, and Classifiers (2024.acl-short)
Copied to clipboard
| Challenge: | Existing evidence supports and challenges the use of isotropy in embedding spaces. |
| Approach: | They propose to formalize this connection mathematically and empirically and prove it's true . they argue that isotropy imposes requirements on embedding space that are not compatible with clusters . |
| Outcome: | The proposed method sheds light on previous studies focusing on anisotropy in embedding spaces. |
Can Out-of-Distribution Evaluations Uncover Reliance on Prediction Shortcuts? A Case Study in Question Answering (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing work assesses models’ generalization capabilities through the lens of performance on out-of-distribution (OOD) datasets. |
| Approach: | They challenge this assumption by comparing OOD evaluations with failure modes documented in existing question-answering (QA) models. |
| Outcome: | The proposed evaluations show that the models' generalization capabilities are under-performing on out-of-distribution datasets, while others are underperforming on in-difference datasets. |
Your Model is Overconfident, and Other Lies We Tell Ourselves (2025.acl-long)
Copied to clipboard
| Challenge: | Analyzing 29 models, we find that difficulty is not linear or monotonic. |
| Approach: | They examine the interplay and divergence among various metrics for assessing intrinsic difficulty, including annotator dissensus, training dynamics, and model confidence. |
| Outcome: | The proposed model is based on 29 models on three datasets and analyzed by a linguistics team. |
What Meaning-Form Correlation Has to Compose With: A Study of MFC on Artificial and Natural Language (2020.coling-main)
Copied to clipboard
| Challenge: | Compositionality is a widely discussed property of natural languages, although its exact definition has been elusive. |
| Approach: | They propose that compositionality can be measured by measuring meaning-form correlation . they analyze three sets of languages: artificial toy languages tailored to be compositional . |
| Outcome: | The proposed method can assess compositionality on three sets of languages . linguistic phenomena such as synonymy and ungrounded stop-words weigh on the results . |
How to Dissect a Muppet: The Structure of Transformer Embedding Spaces (2022.tacl-1)
Copied to clipboard
| Challenge: | Pretrained embeddings based on the Transformer architecture have taken the NLP community by storm . a novel decomposition of Transformer output embeddables is demonstrated . |
| Approach: | They propose to decompose Transformer output embeddings into a sum of vector factors . they show multi-head attentions and feed-forwards are not equally useful in downstream applications . |
| Outcome: | The proposed method outperforms recurrent architectures on a wide variety of tasks. |
Can Machine Translation Bridge Multilingual Pretraining and Cross-lingual Transfer Learning? (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing models that pretrain for cross-lingual tasks do not improve cross-linguistic learning. |
| Approach: | They propose to employ machine translation as a continued training objective to enhance language representation learning by bridging multilingual pretraining and cross-lingual applications. |
| Outcome: | The proposed model performance is compared with existing models and their latent representations. |