35 citations · 59 across the 13 of their papers we have counts for
15 papers
A Polyphonic Conception of AI Understanding
Matthieu Queloz, Pierre Beckmann
When a doctor, a judge, or an engineer must decide whether to trust an AI model's output, they cannot avoid asking what the model understands. Purely mathematical or statistical de…
World Modeling in Transformers
Pierre Beckmann, Matthieu Queloz, Andre Freitas
Behavioral failures can make a transformer appear to lack a world model even when it has learned faithful representations of its environment. We demonstrate this in TaxiGPT, a tran…
Why We Care About Understanding: Competence through Predictive Compression
Matthieu Queloz, Pierre Beckmann
What is the relation between understanding and compression, and why does human understanding take such a heavily compressed form? Across information theory, machine learning, and A…
Deep Learning Models Also Recall Features
Pierre Beckmann
Recent work in mechanistic interpretability has studied how large language models recall facts stored in their weights. This paper argues that factual recall points to something br…
SciR: A Controllable Benchmark for Scientific Reasoning in LLMs
Pierre Beckmann, Marco Valentino, Andre Freitas
Three paradigmatic forms of inference recur across scientific reasoning: deduction, induction, and causal abduction. Reliably evaluating LLMs on these in scientific settings is cur…
Probing Persona-Dependent Preferences in Language Models
Oscar Gilg, Pierre Beckmann, Daniel Paleka +1
Large language models (LLMs) can be said to have preferences: they reliably pick certain tasks and outputs over others, and preferences shaped by post-training and prompting appear…
Where is the Mind? Persona Vectors and LLM Individuation
Pierre Beckmann, Patrick Butlin
The individuation problem for large language models asks which entities associated with them, if any, should be identified as minds. We approach this problem through mechanistic in…
Adaptive LLM-Symbolic Reasoning via Dynamic Logical Solver Composition
Lei Xu, Pierre Beckmann, Marco Valentino +1
Neuro-symbolic NLP methods aim to leverage the complementary strengths of large language models and formal logical solvers. However, current approaches are mostly static in nature,…
Mechanistic Indicators of Understanding in Large Language Models
Pierre Beckmann, Matthieu Queloz
Large language models (LLMs) are often portrayed as merely imitating linguistic patterns without genuine understanding. We argue that recent findings in mechanistic interpretabilit…
Rejecting Cognitivism: Computational Phenomenology for Deep Learning
Pierre Beckmann, Guillaume Köstner, Inês Hipólito
We propose a non-representationalist framework for deep learning relying on a novel method: computational phenomenology, a dialogue between the first-person perspective (relying on…
BYOL-S: Learning Self-supervised Speech Representations by Bootstrapping
Gasser Elbanna, Neil Scheidwasser-Clow, Mikolaj Kegler +3
Methods for extracting audio and speech features have been studied since pioneering work on spectrum analysis decades ago. Recent efforts are guided by the ambition to develop gene…
Hybrid Handcrafted and Learnable Audio Representation for Analysis of Speech Under Cognitive and Physical Load
Gasser Elbanna, Alice Biryukov, Neil Scheidwasser-Clow +5
As a neurophysiological response to threat or adverse conditions, stress can affect cognition, emotion and behaviour with potentially detrimental effects on health in the case of s…
SERAB: A multi-lingual benchmark for speech emotion recognition
Neil Scheidwasser-Clow, Mikolaj Kegler, Pierre Beckmann +1
Recent developments in speech emotion recognition (SER) often leverage deep neural networks (DNNs). Comparing and benchmarking different DNN models can often be tedious due to the…
Word-level Embeddings for Cross-Task Transfer Learning in Speech Processing
Pierre Beckmann, Mikolaj Kegler, Milos Cernak
Recent breakthroughs in deep learning often rely on representation learning and knowledge transfer. In recent years, unsupervised and self-supervised techniques for learning speech…
Deep speech inpainting of time-frequency masks
Mikolaj Kegler, Pierre Beckmann, Milos Cernak
Transient loud intrusions, often occurring in noisy environments, can completely overpower speech signal and lead to an inevitable loss of information. While existing algorithms fo…