Skip to main content
Search

Université de Montréal

5 related articles

Subscribe to updates
  1. arXiv

    Finance agents meet an open world: none of 150 sessions passes, and all 35 the simulated user approved fail a program check

    FinCosmos opens all three uncertainties of financial agent evaluation at once: a simulated investor reveals a hidden strategy only when asked, the agent must source its own Web data and backtest under a data cutoff, and programs verify the delivery instead of matching a reference answer. Across 150 sessions with four coding agents and a tool-free model, none is accepted, and all 35 sessions the simulated user signed off fail at least one program veto.

  2. arXiv

    Clean Cuts Second-Order Optimizer Memory to Linear via Nyström Sketching, Enabling 13B Pre-training on a Single 80GB GPU

    The authors introduce Clean and its low-precision variant Q-Clean: randomized Nyström approximations replace SOAP's explicit left and right preconditioners, reducing optimizer-state memory from quadratic to linear in layer dimensions, while four orthogonal blocks (Main, A, B, C) reintegrate off-subspace components into the update; pre-training LLaMA-350M and 1.3B on C4, Clean reaches perplexity close to SOAP with less optimizer memory than AdamW, Q-Clean cuts optimizer memory by over 50% versus Muon, Clean reaches AdamW's final performance 26% faster in wall-clock time, and the methods uniquely enable pre-training a 13B model on a single 80GB H100.

  3. arXiv

    LMOPD uses lexicographic multi-teacher distillation to preserve top-priority capabilities at about 90% retention on a 30B-A3B model

    The work introduces Lexicographic Multi-Objective On-Policy Distillation (LMOPD), which for each student rollout selects the specialist for the first objective whose gate detects a deficiency and locally projects its centered log-policy correction to remove components opposing higher-priority specialists; on three math benchmarks with 30B-A3B mixture-of-experts models, the two-expert setting fully retains accuracy and reasoning-quality gains in point estimates while acquiring 46.9% of the conciseness gain, and the four-expert setting retains about 90% of both the accuracy gain and reasoning-correctness gain versus only about 57% for the next best evaluated baseline.

  4. The AWARE Framework: An Educational Interview Framework for Assessing Patients' Use of Conversational AI in Mental Health Care

    This article introduces the AWARE (AI use, why, attachment, reality and risk, and effect on functioning) framework as an educational tool to help mental health professionals systematically assess patients' use of conversational AI through five clinically relevant domains, and discusses incorporating it into undergraduate, postgraduate, and continuing professional education along with priorities for future research.

  5. arXiv

    Re-evaluating the Advancements of Heterophilic Graph Learning: A Three-Way Dataset Taxonomy and Quantitative Evaluation of Homophily Metrics

    This work fine-tunes baseline models on 27 widely used benchmark datasets, categorizes heterophilic datasets into malignant, benign, and ambiguous groups based on whether graph-aware models underperform their coupled graph-agnostic counterparts, re-evaluates 11 SOTA heterophily-specific models, and conducts the first quantitative evaluation of 11 homophily metrics on synthetic graphs from three generation methods, finding that most SOTA models do not significantly outperform the best baselines and that classic metrics remain competitive.

You’re all caught up