Eric Elmoznino

eric.elmoznino@mila.quebec

Doctorat - UdeM

Superviseur⋅e principal⋅e

Yoshua Bengio

Co-supervisor

Guillaume Lajoie

Sujets de recherche

Apprentissage profond

Neurosciences computationnelles

Site web

Google Scholar

GitHub

Publications

Amortizing Intractable Inference in Large Language Models

Edward J. Hu

Nikolay Malkin

Autoregressive large language models (LLMs) compress knowledge from their training data through next-token conditional distributions. This l… (voir plus)imits tractable querying of this knowledge to start-to-end autoregressive sampling. However, many tasks of interest -- including sequence continuation, infilling, and other forms of constrained generation -- involve sampling from intractable posterior distributions. We address this limitation by using amortized Bayesian inference to sample from these intractable posteriors. Such amortization is algorithmically achieved by fine-tuning LLMs via diversity-seeking reinforcement learning algorithms: generative flow networks (GFlowNets). We empirically demonstrate that this distribution-matching paradigm of LLM fine-tuning can serve as an effective alternative to maximum-likelihood training and reward-maximizing policy optimization. As an important application, we interpret chain-of-thought reasoning as a latent variable modeling problem and demonstrate that our approach enables data-efficient adaptation of LLMs to tasks that require multi-step rationalization and tool use.

2024-05-06

International Conference on Learning Representations (Accept (oral))

doi.org

openreview.net

Explicit Knowledge Factorization Meets In-Context Learning: What Do We Gain?

2024-03-04

ICLR.cc/2024/Workshop/R2-FM (poster)

openreview.net

Discrete, compositional, and symbolic representations through attractor dynamics

Andrew Nam

Eric Elmoznino

Nikolay Malkin

Chen Sun

Yoshua Bengio

Guillaume Lajoie

Compositionality is an important feature of discrete symbolic systems, such as language and programs, as it enables them to have infinite ca… (voir plus)pacity despite a finite symbol set. It serves as a useful abstraction for reasoning in both cognitive science and in AI, yet the interface between continuous and symbolic processing is often imposed by fiat at the algorithmic level, such as by means of quantization or a softmax sampling step. In this work, we explore how discretization could be implemented in a more neurally plausible manner through the modeling of attractor dynamics that partition the continuous representation space into basins that correspond to sequences of symbols. Building on established work in attractor networks and introducing novel training methods, we show that imposing structure in the symbolic space can produce compositionality in the attractor-supported representation space of rich sensory inputs. Lastly, we argue that our model exhibits the process of an information bottleneck that is thought to play a role in conscious experience, decomposing the rich information of a sensory input into stable components encoding symbolic information.

2023-10-26

NeurIPS.cc/2023/Workshop/InfoCog (présentation orale)

doi.org

openreview.net

Learning Macro Variables with Auto-encoders

Dhanya Sridhar

Eric Elmoznino

Maitreyi Swaroop

2023-10-26

NeurIPS.cc/2023/Workshop/CRL (poster)

openreview.net