Jean-Pierre Falet

jean-pierre.falet@mila.quebec

Doctorat - UdeM

Superviseur⋅e principal⋅e

Yoshua Bengio

Sujets de recherche

Apprentissage profond

Apprentissage sur graphes

Modèles probabilistes

Raisonnement

Sécurité de l'IA

Site web

Google Scholar

GitHub

Publications

A longitudinal study of research productivity in McGill University's MD-PhD and Clinician Investigator Program Graduates from 2000 to 2017

Jamie Magrill

Jean-Pierre Falet

Ella Boone

Mimosa Luigi

Isabella Arthur

Paria Asadi

Anna Frumkin

Mark Sorin

Joan Miguel Romero

Julia Luo

Nathalie Johnson

Mark J. Eisenberg

BACKGROUND: Canadian MD-PhD and Clinician Investigator Program (CIP) pathways both train physician-scientists, yet longitudinal data on grad… (voir plus)uate research productivity are limited. METHODS: We conducted a longitudinal study of McGill University MD-PhD and CIP graduates who completed these programs between 2000 and 2017, using sex- and year-matched MD-only graduates as controls. Outcomes were collected for publications indexed through to December 31, 2022, and included H-index, total publications, total citations, per-paper journal impact factor, per-paper citations, and authorship position (first, second, senior). Productivity was assessed across four time periods: (1) pre-medical/medical training, (2) residency, (3) post-residency/fellowship, and (4) independent practice. RESULTS: Among 549 graduates (MD-PhD = 31; CIP = 97; MD-only = 421), research productivity, research impact, and active research involvement (defined as ≥3 first- or senior-author papers in the prior 5 years) were similar between MD-PhD and CIP graduates; both exceeded MD-only graduates across all metrics. Within-group sex differences were not significant. MD-PhD graduates were more productive during pre-medical/medical training, whereas CIP graduates were comparatively more productive during residency. Sustained productivity in independent practice correlated with research engagement during medical school, residency, and fellowship. In both programs, graduates with active research involvement had higher authorship counts in the 10 years immediately following graduation from medical school. DISCUSSION: Research productivity and impact were comparable between MD-PhD and CIP graduates, and both groups exceeded MD-only peers across measured research metrics. Across both programs, graduates with active research involvement had higher early-career authorship counts, particularly during residency and post-residency/fellowship time periods. These findings provide descriptive benchmarking data for future studies of physician-scientist training pathways in Canada.

2026-06-30

Clinical and Investigative Medicine (publié)

doi.org

Latent Veracity Inference for Identifying Errors in Stepwise Reasoning

Oliver E. Richardson

Sungsoo Ahn

Chain-of-Thought (CoT) reasoning has advanced the capabilities and transparency of language models (LMs); however, reasoning chains can cont… (voir plus)ain inaccurate statements that reduce performance and trustworthiness. To address this, we propose to augment each reasoning step in a CoT with a latent veracity (or correctness) variable. To efficiently explore this expanded space, we introduce Veracity Search (VS), a discrete search algorithm over veracity assignments. It performs otherwise intractable inference in the posterior distribution over latent veracity values by leveraging the LM's joint likelihood over veracity and the final answer as a proxy reward. This efficient inference-time verification method facilitates supervised fine-tuning of an Amortized Veracity Inference (AVI) machine by providing pseudo-labels for veracity. AVI generalizes VS, enabling accurate zero-shot veracity inference in novel contexts. Empirical results demonstrate that VS reliably identifies errors in logical (ProntoQA), mathematical (GSM8K), and commonsense (CommonsenseQA) reasoning benchmarks, with AVI achieving comparable zero-shot accuracy. Finally, we demonstrate the utility of latent veracity inference for providing feedback during self-correction and self-improvement.

2026-01-25

International Conference on Learning Representations (poster)

doi.org

openreview.net

Delta-AI: Local Objectives for Amortized Inference in Sparse Graphical Models

Jean-Pierre Falet

Hae Beom Lee

Nikolay Malkin

Chen Sun

Dragos Secrieru

Dinghuai Zhang

Guillaume Lajoie