The upcoming meeting, taking place on November 10 at Mila, will explore how we can collectively develop, govern, and deploy high-performing, reliable, and secure agentic systems by connecting academic researchers, industry experts, and practitioners.
Join us on the Venture Scientist Bootcamp, a full time, 4-month incubator at Mila, built specifically for deep tech founders with elite STEM backgrounds. You have until November 21, 2026 to apply!
We use cookies to analyze the browsing and usage of our website and to personalize your experience. You can disable these technologies at any time, but this may limit certain functionalities of the site. Read our Privacy Policy for more information.
Setting cookies
You can enable and disable the types of cookies you wish to accept. However certain choices you make could affect the services offered on our sites (e.g. suggestions, personalised ads, etc.).
Essential cookies
These cookies are necessary for the operation of the site and cannot be deactivated. (Still active)
Analytics cookies
Do you accept the use of cookies to measure the audience of our sites?
Multimedia Player
Do you accept the use of cookies to display and allow you to watch the video content hosted by our partners (YouTube, etc.)?
Publications
Synthetic Content Validation of Pediatric Trust Instruments Using Persona-Driven Large Language Models: Development and Evaluation Study
Background: Large language models (LLMs) could assist in streamlining health care instrument validation. This is particularly relevant as tr… (see more)aditional instrument development is time-intensive and resource-intensive. Currently, there is a significant gap in specifically developed and validated trust instruments for pediatric emergency and surgical contexts. Such tools are key because trust is foundational to the patient family-physician relationship and is associated with improved care-seeking and treatment adherence. Objective: This study aimed (1) to develop new trust instruments and conduct their preliminary content validation confirming item relevance through a Synthetic Instrument Validation approach integrating human and LLM capabilities and (2) to evaluate appropriate use cases for LLMs in psychometric assessment. Methods: Two trust instruments were developed (patient family and physician versions). In phase 1, instruments underwent a 2-stage content validation process using parallel synthetic and human expert panels (5 patient family members, 3 physician synthetic experts, and 7 unique human experts). Synthetic panels consisted of 3 persona-prompted LLMs (Claude Sonnet 4, GPT-5, and Grok 4), with human panels serving as comparators. The scale-content validity index (S-CVI) threshold was set at ≥0.80, percent agreement on item relevancy was used to compare panels. In phase 2, LLM performance in quantitative research tasks was evaluated. The patient family instrument underwent Flesch-Kincaid assessment; both instruments underwent cosine similarity analyses using algorithmic and LLM-based methods. Results: In phase 1, percent agreement between panels for the patient family and physician instruments was 90.9% and 88.9% for dimensional validation and 90.9% and 100% for contextual validation. Fleiss κ for within-panel agreement ranged from poor to almost perfect; 0.20 (95% CI -0.23 to 0.63) to 1.00 (95% CI 1.00-1.00). All instrument subsections met or exceeded S-CVI thresholds across both rounds and panels. Phase 2 Flesch-Kincaid metrics converged across all 3 methods to target benchmark (mean Grade Level 8.1, SD 1.1; mean Reading Ease 60.1, SD 5.6), while reference readability was near target, showing convergence across implementations of the same formula. In contrast, cosine similarity analyses revealed significant LLM performance limitations; cosine similarity analyses were, therefore, abandoned in favor of the algorithmic method alone, which achieved a maximum cosine similarity of 0.83, indicating acceptable item distinctiveness overall. Conclusions: Persona-prompted LLMs effectively assisted preliminary content validation and may help reduce timelines from months to weeks, but showed limitations in quantitative computations. This suggests that, in this study, persona-prompted LLMs performed better in subjective rating tasks than in deterministic cosine similarity computations. These findings help illustrate task-dependent boundaries for LLM integration in psychometric research, necessitating selective human-LLM collaboration. While these instruments achieved preliminary content validity, they require full psychometric testing before they can be considered validated for clinical use. Synthetic Instrument Validation offers a feasible approach to preliminary health care instrument development while preserving key elements of content validation.
Buildings cause nearly 40\% of carbon emissions in the US. Of total building emissions, HVAC systems are responsible for approximately 40-60… (see more)\% and account for 15\% of global energy consumption. Despite advances in artificial intelligence, most buildings still operate as inefficiently as they did decades ago. Building optimization, perhaps more so than any other optimization problem, has enormous potential for Machine Learning (ML) driven impact on climate change, in both the short and long term. One of the biggest obstacles for building optimization is the lack of policy transfer. Most building control systems, whether simulation or learning based, don't generalize: they only work well for the building they were developed for, but not for other buildings. To solve this problem, we propose a challenge for the development of building foundation models that easily adapt to new buildings with minimal configuration and thereby enable building optimization to scale across the globe. Participants will build control policies on a set of simulated buildings, and have their policies evaluated on an unseen set of buildings. The buildings vary in size, layout, weather, material properties, and HVAC system. Successful outcomes from this initiative promise a massive leap forward in global sustainability efforts and demonstrate the great potential of ML to combat climate change.
Long wait times for colonoscopy and challenges in referral prioritization contribute to procedural backlogs in endoscopy units, potentially … (see more)delaying the diagnosis of colorectal cancer (CRC) and other clinically significant conditions. In Québec, colonoscopy referrals are currently prioritized using structured clinical indications documented in the AH-702 referral form. This study aimed to develop and externally validate an explainable artificial intelligence (AI) framework for predicting colonoscopy referral priorities across multiple healthcare centers while addressing missing data and class imbalance. This multicenter retrospective study included 11,680 patients referred for colonoscopy at two tertiary centers in Québec, Canada, between 2018 and 2022. Using 135 clinical, demographic, diagnostic, and pathological features, a multiclass decision-tree model was trained with advanced imputation and synthetic data generation. The model was evaluated on an internal held-out test set and an independent external cohort of 2,977 patients. The optimized model achieved an area under the receiver operating characteristic curve (AUROC) of 98.4% and a balanced accuracy of 95.7% on the internal held-out test set. External evaluation yielded an AUROC of 97.9% and a balanced accuracy of 96.4%. When the 22 structured AH-702 indication fields were withheld, balanced accuracy decreased to 28.1% internally and 28.5% externally, while AUROC decreased to 60.1% and 61.3%, respectively. These findings demonstrate that the model identified clinically relevant referral indications as key predictors, consistent with the existing AH-702 prioritization framework. This interpretable AI framework demonstrated strong retrospective performance for predicting colonoscopy referral priorities across three healthcare centers. The findings support further investigation of its potential as a decision-support tool for referral prioritization, pending prospective validation.
2026-10-11
International Journal of Colorectal Disease (published)
Video-to-robot pipelines for quadrupedal imitation insert a learned lifting stage, either a monocular depth model or a per-clip 3D reconstru… (see more)ction, to recover the one axis a side-view camera cannot see. We present CopyQuad, which omits that stage and pins the missing axis to the robot's own stance width instead. The setting is side-view footage of sagittally dominated motion, where limb motion is almost entirely in the image plane. A frozen 2D keypoint detector and a fixed analytic retargeting then turn a single uncalibrated clip into a robot reference whose foot targets are reachable by construction, so no per-clip trajectory optimization is needed. One phase-indexed reward structure, whose only per-behavior freedom is three scalars from a shared sweep, learns a dog walk, a horse walk, a horse trot and two non-periodic gestures, all deployed zero-shot on hardware. We then measure what omitting the axis costs. In a training-free rigidity test, a video depth model varies the distance between the animal's hips by
In this work, we explore the idea of expert intervention delays in human-supervised robotic systems for imitation learning system. Previous … (see more)works in imitation learning have primarily examined the impact of suboptimal experts, where the expert policy makes mistakes, but assumed expert interventions occur immediately after errors are made. Although delayed expert intervention can be viewed as a form of noise that might improve policy robustness, our results show that learning from delayed expert interventions is overall less effective than learning from non-delayed expert interventions; delayed expert interventions are less effective at demonstrating error avoidance behaviors. To address this, we propose Learning from Delayed Intervention (LfDI), which learns a cost function based on where and when the expert intervenes, combined with limited information about the extent of intervention delays provided by the expert. This cost function is then used to infer additional supervision for states without expert data, guiding the policy to avoid states that would likely trigger interventions. We evaluate our method, Learning from Delayed Intervention (LfDI), on simulated environments and show that it outperforms other methods that do not account for intervention delays.
Antibody therapeutics have transformed treatment across diverse therapeutic areas and have become central to modern medicine. While recent m… (see more)achine learning approaches aim to accelerate antibody discovery through structure-based design and target-affinity optimization, successful candidates must also satisfy broad developability constraints, including structural compatibility, low immunogenicity, and favorable biophysical properties, while remaining sufficiently diverse to provide sufficient coverage of design space to mitigate later-stage development failures. We introduce a preference-conditioned multi-objective generative flow network (GFN) fine-tuning pipeline for fixed-framework CDRH3 loop redesign that fine-tunes a biological sequence model against many inexpensive proxies covering backbone compatibility, sequence likelihood, immunogenicity risk, novelty, and developability liabilities. It combines up to fourteen predictive rewards in a single policy trained to sample CDRH3 loops across 623 SAbDab antibody structures, leveraging a reward reduction strategy to preserve Pareto dominance under this learned map. We also introduce a novel autoregressive span-in-filling variant of AMPLIFY-350M that serves both as a scaffold-conditioned CDRH3 generation prior and as a direct conditional likelihood scorer. We evaluated reward scores, diversity, and design quality on loops sampled on 45 held-out antibody complexes to assess structural consistency, confidence, aggregation propensity, stability, and antigen binding. Compared with the inverse-folding methods ProteinMPNN and ESM-IF, our approach produces roughly 10x more distinct acceptable candidates per design at 50% sequence identity, while holding 10% of the 14-objective Pareto front pooled across methods, trading per-sample quality for greater exploration of novel possible structures.
Diffusion language models are a promising and more efficient alternative to autoregressive models. However, most RL post-training techniques… (see more) are not easily adapted to these models. One issue is that diffusion language models generate answers through many intermediate states, deciding at each step which masked positions to reveal and which tokens to place in them. Many existing RL methods used in this setting approximate intractable sequence likelihoods or weight every denoising step by the same advantage, without explicitly distinguishing how these intermediate states of the denoising trajectory affect the expected reward. To address this issue of trajectory blindness, we introduce PACT (Potential-Anchored Consistency of Transitions). Motivated by a reward-tilted reference distribution, PACT jointly learns the policy and a scalar potential that scores each intermediate state, anchored by terminal reward. For each sampled denoising transition, a consistency loss matches the policy-to-reference log-probability ratio of revealed tokens to the change in this potential. This formulation requires neither sequence-likelihood estimation nor importance-ratio clipping. On four tasks, PACT improves LLaDA-8B-Instruct by +9.9 points on GSM8K, +12.0 on MATH500, +55.7 on Countdown, and +72.6 on Sudoku, and reaches the best average accuracy across the four tasks reported for this model, +3.7 points over the next best method. Further compared to AGRPO, it reaches that baseline's best accuracy in about half the training compute on GSM8K.
Recursive reasoning models such as the Tiny Recursive Model (TRM) iteratively refine two latent states: an answer state y, decoded into the … (see more)output, and a reasoning state z, which is never decoded. Many of their failures have been attributed to bad basins, regions of latent space that trap the recursion on wrong answers. Recent test-time approaches perturb y, z or both to recover these failures, but the impact of this choice on performance, test-time compute efficiency and scaling properties remains unclear. Through controlled experiments on TRM, we find that a failed trajectory carries its failure in its answer state. At equal compute, whether retries continue a failed trajectory (depth scaling) or draw multiple trajectories from the initial states (width scaling), resampling y recovers more failures than any perturbation of z we tested, and the performance gap widens as the test-time compute budget grows. Our approach, Gated Answer Resampling (GAR), pairs this resampling of the answer state with the model's own halting head and restarts only the puzzles that the head rejects. With at most 64 restarts per puzzle and no training or hyperparameter search, GAR takes TRM from 85.9% to 99.4% on Sudoku-Extreme and from 68.8% to 85.6% on Pencil Puzzle Bench validation. On the same checkpoints, GAR also outperforms Probabilistic TRM (PTRM), which perturbs z at each step, even when PTRM uses 42 to 197 times as much compute. Our code is available at https://github.com/MaximeILL/GAR.
Medical Concept Recognition is the task of extracting and grounding spans in medical text to entries in a standard ontology. Current methods… (see more) are generally developed towards a single ontology and source text type (clinical notes, drug labels, etc.) The central challenge is therefore to build methods that adapt to any target ontology without requiring ontology-specific retraining. More recent approaches like REAL and AutoPCR make direct use of Large Language Models and can theoretically adapt to any ontology choice but suffer from a lack of generalization. We propose UMCR, a pipeline that allows extraction and linking to any target ontology using any source text variety, utilizing direct guideline integration and an agentic linking module. We evaluate our method on the recent MedPath benchmark, covering a wide variety of medical text types and vocabularies, and show that our method achieves strong performance across these tasks. This ability unlocks interoperability in the healthcare domain, allowing for extraction and alignment between different medical standards. Our method also reduces dependence on parametric LLM memory, thanks to improved interaction with the target ontology compared to other proposed approaches.
Abstract The upper cervical spinal cord is essential for motor, sensory and autonomic function, yet its normative morphometric trajectories … (see more)and disease-related deviations remain unclear. Using three-dimensional T1-weighted images from 10,159 healthy individuals aged 8-85 years, we established normative references for averaged and level-specific upper cervical spinal cord morphometry, revealing measure-specific trajectories with peak ages of 29.7-44.1 years. These references were applied to 2,394 patients with seven neurological disorders: mild cognitive impairment, Alzheimer’s disease, Parkinson’s disease, cerebral small vessel disease, multiple sclerosis, neuromyelitis optica spectrum disorder and myelin oligodendrocyte glycoprotein antibody-associated disease to derive individualized deviation scores. These deviation scores captured disease-specific abnormalities, enabled disease classification, and were associated with cognition, physical disability and disease progression. Finally, a personalized upper cervical spinal cord report was developed to support individualized imaging interpretation. Together, these findings provide a normative framework for quantifying upper cervical spinal cord morphometry and advancing personalized assessment across neurological disorders.
Modern neural-network training uses mixed recipes, assigning AdamW-like or sign-based updates to embedding and output layers and spectral up… (see more)dates to hidden matrices. These assignments are usually fixed by layer type, without a clear criterion for selecting the appropriate geometry. We turn sign-versus-spectral assignment into a measurable, layerwise decision problem. For ideal memoryless steps, we derive a finite-batch ratio of conservative descent-lemma lower bounds for spectral and sign updates. Its closed form separates signal efficiency from a geometry-dependent global curvature factor. In the positive-signal regime, the same smoothness assumption gives two-sided bounds that determine which step has the larger conditional expected decrease when the ratio crosses explicit noise-dependent thresholds. We then derive a local directional quadratic score from gradient alignments and Hessian-vector products and use it in Generalized-Scion (G-Scion) to control layerwise routing. In a controlled input-rank experiment, including directional curvature matches the measured same-iterate ordering in all seven aggregate settings, whereas the curvature-free proxy selects the opposite geometry. On 31M GPT-2, the output score routes the output layer from spectral warmup to sign, and G-Scion improves validation loss over fixed Scion in memoryless and matched-momentum comparisons. ViT-B/16 results further show that a spectral edge assignment can outperform fixed Scion outside language modelling. These results suggest that optimizer geometry need not be hard-coded by layer type and are a first step toward understanding how it should be assigned.