Publications

Do LLMs Beat Nash? Testing Decentralized Coordination in Self-Play Multi-Agent Games
Deborah Sinishaw
Qile Zhu
Edwin Meriaux
Gregory Dudek
Large language model agents deployed without a central controller are often assumed to require communication to coordinate their actions. We… (see more) ask what remains possible without it: when independent instances of the same model cannot communicate, can they still reason about their counterparts well enough to exceed the standard game-theoretic baseline for uncoordinated play? We introduce a benchmark of one-shot, no-communication games in which each of thirteen language models is told only that its counterparts are running the same model and is evaluated against the Nash equilibrium of the underlying game. In two-player matrix games spanning seven archetypes and two to ten actions per player, two frontier-hosted models consistently exceed their Nash benchmark, approaching the optimal joint outcome in several archetypes, while most open-weight models achieve only partial gains that vary sharply by game structure. Performance degrades substantially in team-based games with four or more interchangeable agents, particularly as the action space grows, suggesting that whatever capability drives self-play gains in dyadic games does not transfer to larger multi-agent teams.
2026-08-11
arXiv (preprint)
doi.org
arxiv.org
Backdoor Decontamination Dynamics in LLM Agents
Gabriel Huang
Abhay Puri
Léo Boisvert
Alexandre Drouin
Perouz Taslakian
Spandana Gella
Christopher Pal
Open-weight LLM agents are vulnerable to backdoors installed during fine-tuning, which may be undetectable if the trigger conditions are nev… (see more)er met during testing. Assuming defenders do not know the existing trigger, they cannot unlearn it directly. One decontamination strategy is to install a known backdoor (defensive poisoning) then to unlearn it, hoping that the original unknown backdoor is removed as a side effect. However, this procedure has uncertain outcomes: the original backdoor may persist or be erased or rerouted, among other possibilities. We introduce a framework for studying these dynamics in tool-calling agents, decoupling trigger, response, teacher, and fine-tuning method across systematic experiments on AgentDyn. Across 115 experiments, defensive poisoning alone erases around 56% of original backdoors; subsequent decontamination then drives almost all survivors to erasure, confirming that trigger recognition and malicious execution are behaviorally dissociable. Interestingly, our experiments find that malicious backdoors never persist when using different triggers of the same general type as the defensive backdoor when followed by decontamination via unlearning. Co-installing up to four backdoors increases resistance (around 36% erased), yet decontaminating a single known co-resident backdoor collaterally clears 52/60 co-residents (87%). Upon visualizing postdecontamination model internals using J-lens, we confirm that although the decontamination restores benign LLM responses, traces of original trigger awareness persist at intermediate layers.
2026-08-10
arXiv (preprint)
doi.org
arxiv.org
Generative Learning for Quantum Measurement Design
Jun Dai
Olivier Nahman-Lévesque
Guillaume Rabusseau
Hong-Ye Hu
Cunlu Zhou
Extracting quantum information from a quantum state is a fundamental task of quantum computation, often requiring the estimation of many non… (see more)-commuting observables under a finite measurement budget. For both near-term and early fault-tolerant settings, the measurement protocol must balance statistical efficiency against implementation resources such as circuit depth, connectivity, and entangling-gate count. Many existing strategies focus on two extremes: hardware-friendly product measurements with high sampling cost, and fully commuting measurements with deep circuits. Here we recast resource-constrained measurement design as a generative learning problem. We introduce FlowMeas, which uses a generative flow network to directly sample finite ensembles of shallow Clifford measurement circuits subject to a prescribed shot budget and hardware constraints. At zero entangling depth, FlowMeas learns qubit-wise commuting measurement schedules and already matches or improves leading product-measurement methods on nearly all molecular benchmarks. Allowing one or two entangling gate layers yields further reductions in energy estimation error of up to
2026-08-10
arXiv (preprint)
doi.org
arxiv.org
Koopman Representation of Nonlinear Virtual Environments in Kinesthetic Haptic Systems
Yanting Zhou
József Kövecses
James Richard Forbes
Rendering haptic feedback with nonlinear virtual environments (VEs) is important in many applications that require highly accurate force fee… (see more)dback. This paper considers the use of the Koopman operator to represent a nonlinear VE interacting with a haptic system. Simulation and experimental results demonstrated that the proposed method provides an effective representation of the nonlinear dynamics of a Duffing-oscillator VE. A multi-user study further confirmed this conclusion. In addition, a closed-loop (CL) stability analysis is performed leveraging the Koopman representation of the nonlinear VE to access stability of the overall haptic system. This alternative way of representing nonlinear VEs enables a convenient CL stability analysis that is less conservative than traditional passivity-based methods. Since a linear combination of all lifted states is used to represent the nonlinearity, such representation is also more robust to uncertainties in the modeling of the haptic device than a traditional nonlinear model.
2026-08-10
arXiv (preprint)
doi.org
arxiv.org
VoxSumm: A Multilingual Corpus of Long-Form Spoken News for Joint Summarization and Translation
Yejin Jeon
Marie Maltais
Virginia Ceccatelli
Min Ma
David Ifeoluwa Adelani
As information increasingly traverses linguistic boundaries, users require concise cross-lingual representations of long-form content. Never… (see more)theless, long-document summarization research remains text-centric, whereas multilingual speech research has largely prioritized translation, preserving source content rather than compressing it. We address this methodological gap by formalizing joint speech summarization and translation (JSumT): the generation of a succinct, faithful target-language summary directly from a long spoken document in a source language. We additionally introduce VoxSumm, the first multilingual and cross-lingual benchmark for this task, comprising 10,045 BBC article-summary pairs across 24 languages and encompassing approximately 703 hours of speech data. Our evaluation of representative speech-language models reveals pronounced variation across models and generation settings: Gemini3.1-Pro demonstrates the greatest consistency, summarization into English generally surpasses generation into non-English target languages, and translating an entire document before summarization compounds instruction-following failures. Through the release of VoxSumm, we establish a foundation for developing and evaluating multilingual systems capable of jointly interpreting, compressing, and translating long-form speech.
2026-08-10
arXiv (preprint)
doi.org
arxiv.org
Assessing the effect of race on machine learning models predicting hospital admissions from emergency department
Aidan Licoppe
Lucy An
David L. Buckeridge
Antony Robert
James Man Git Tsui
Abstract Objectives Machine learning (ML) models are increasingly being developed to support healthcare delivery. However, concerns remain a… (see more)bout their potential to perpetuate existing biases rooted in the data used to develop them. We aim to assess the impact of using race in predicting hospital admission probabilities for patients visiting the emergency department (ED). Materials and Methods Data from the MIMIC-IV ED dataset were used to train2 ML models predicting hospital admission: one included race; the other did not. Differences in predicted admission probabilities were evaluated across racial groups under multiple validation conditions. Results Including race as a model input was associated with meaningful differences in predicted admission probabilities for White (3.2%), Black (−1.5%), and Hispanic (−3.0%) patients, while minimal differences were observed for Asian (0.2%) and Other (0.5%) patients. These differences were associated with large Cohen’s d effect sizes in the baseline model for White (d = 1.00), Black (d = −1.23), and Hispanic (d = −1.35) patients. After balancing racial group prevalence, the effects persisted for White (1.22%; d = 1.22) and Hispanic (−1.00%; d = −1.00) patients. Discussion These findings suggest that race is associated with differences in ML-based hospital admission predictions for ED patients, underscoring the need for caution when incorporating race into clinical prediction models and the importance of rigorous bias assessment. Conclusion As race was associated with predictions, there is a crucial need to address underlying social factors and the use of broader, more equitable clinical data for ML model training.
2026-08-09
Journal of the American Medical Informatics Association Open (published)
doi.org
Invariant Causal Set Covering Machine
Thibaud Godon
Baptiste Bauvin
Pascal Germain
Jacques Corbeil
Alexandre Drouin
Rule-based models, such as decision trees, appeal to practitioners due to their interpretable nature. However, the learning algorithms that … (see more)produce such models are often vulnerable to spurious associations, and thus, they are not guaranteed to extract causally relevant insights. This limitation reduces their utility in gaining mechanistic insights into a phenomenon of interest. In this work, we build on ideas from the invariant causal prediction literature to propose Invariant Causal Set Covering Machines, an extension of the classical Set Covering Machine (SCM) algorithm for conjunctions/disjunctions of binary-valued rules that provably avoids spurious associations. The proposed method leverages structural assumptions about the functional form of such models, enabling an algorithm that identifies the causal parents of a variable of interest in polynomial time. We demonstrate the validity and efficiency of our approach through a simulation study and highlight its favorable performance compared to SCM in uncovering causal variables across real-world datasets.
2026-08-09
Transactions on Machine Learning Research (accepted)
openreview.net
openreview.net
CrediBench: Building Web-Scale Network Datasets for Information Integrity
Emma Kondrup
Sebastian Sabry
Hussein Abdallah
Zachary Yang
James Zhou
Kellin Pelrine
Jean-François Godbout
Michael M. Bronstein
Reihaneh Rabbany
Shenyang Huang
Online misinformation poses an escalating threat, amplified by the Internet's open nature and increasingly capable LLMs that generate persua… (see more)sive yet deceptive content. Existing misinformation detection methods typically focus on either textual content or network structure in isolation, failing to leverage the rich, dynamic interplay between website content and hyperlink relationships that characterizes real-world misinformation ecosystems. We introduce CrediBench: a large-scale data processing pipeline for constructing temporal web graphs that jointly model textual content and hyperlink structure for misinformation detection. Unlike prior work, our approach captures the dynamic evolution of general misinformation domains, including changes in both content and inter-site references over time. Our processed one-month snapshot extracted from the Common Crawl archive in December 2024 contains 45 million nodes and 1 billion edges, representing the largest web graph dataset made publicly available for misinformation research to date. From our experiments on this graph snapshot, we demonstrate the strength of both structural and webpage content signals for learning credibility scores, which measure source reliability. The pipeline and experimentation code are all available here, and the dataset is in this folder.
2026-08-07
ACM SIGKDD Conference on Knowledge Discovery and Data Mining (published)
doi.org
openreview.net
What Survives KV-Cache Compression for Reasoning? A Dual-Channel View of Structural and Associative Memory
Md Rifat Arefin
Irina Rish
Long reasoning traces make Transformer key–value (KV) caches grow linearly with generation length. Existing compression methods adopt diff… (see more)erent memory topologies, but it is unclear how those choices affect the information that survives eviction. We observe a task–fidelity mismatch: dense summaries achieve low KV reconstruction error yet lose information required for structured reasoning. We explain this through a dual-channel model of compressed memory consisting of a temporally organized scaffold and sparse associative bindings, suggesting that structure and association should be stored separately. We instantiate this idea as TT–Delta, which combines a Tensor-Train structural state with a Delta-rule associative state. On a frozen Qwen2.5-0.5B task model, TT–Delta compresses an evicted prefix by 25.3× (46,680 persistent scalars) and improves exact-answer generation on sequential linear-system reasoning from 0.302±0.113 to 0.839±0.171. In contrast, multi-head Delta is strongest on associative recall and branching graph search. These results suggest that no single compressed-memory topology is universally optimal; the appropriate topology depends on the structure of the reasoning state.
2026-08-06
Efficient Reasoning @ Conference on Language Modeling (published)
openreview.net
openreview.net
A Passivity-Based Analysis of First-Order Momentum-Based Methods
Sepehr Moalemi
James Richard Forbes
This paper presents a discrete-time passivity-based analysis of first-order momentum-based methods for a class of functions whose gradient h… (see more)as lower and upper sector bounds of
2026-08-05
arXiv (preprint)
doi.org
arxiv.org
VIDP: Variable Impedance Diffusion Policy for Compliant Robot Manipulation from Diverse Demonstrations
Hisham Khalil
Neil Fernandes
Thomas M. Kwok
Hsiu-Chin Lin
Yue Hu
Contact-rich manipulation requires precise tracking and mechanical compliance, where variable impedance control can improve robustness in ta… (see more)sk success, whereas static compliance cannot adapt to varying contact constraints. Variable impedance skills can be learned from demonstrations, avoiding complex modeling, but compliance is a hidden variable in force-agnostic kinematic data. While existing methods infer compliance from trajectory variations, these variations may reflect geometric adaptation and not intentional compliance when subject to changing spatial layouts. Therefore, this letter introduces Variable Impedance Diffusion Policy (VIDP), an imitation learning-based variable impedance control framework leveraging a Task-Parameterized Directionality-Aware Mixture Model (TP-DAMM) to extract physically consistent trajectory distributions from diverse demonstrations. By mapping distributions to stiffness profiles, VIDP jointly predicts pose actions and task compliance without force sensors. Real-world experiments show that VIDP significantly outperforms fixed-impedance baselines in task success rate while reducing interaction forces with respect to high stiffness controllers and tracking errors with respect to low stiffness baselines.
2026-08-05
arXiv (preprint)
doi.org
arxiv.org
A Chain Is Only as Strong as Its Weakest Link: A Scoping Review of System Integration Audits in AI
Leah Davis
Dominic Martin
AJung Moon
As AI systems become increasingly integrated into diverse interfaces and applications, model-centric audits are insufficient to address risk… (see more)s arising from interactions among system components and deployment environments. System integration has long been central to software audits in safety-critical domains such as aerospace. However, its role in AI auditing remains underexplored. Scanning through 4,259 documents, we present a scoping review of AI audits that treat system integration as a core tenet of evaluation (n = 58). Using reflexive thematic analysis, we analyze their elements, actors, enablers, and constraints. We find that the corpus represents an emerging yet still fragmented form of AI auditing: few existing measures target integration-specific risks; large gaps remain in meeting traditional audit expectations; and access to necessary information and resources significantly influences audit design. Nonetheless, integration can be categorized across three sites (inter-component, system-environment, and multi-system), each serving the functions of risk exploration, risk determination, coordination, and procedural regularity. Deviating from other types of evaluations, these audits assess qualities specific to system integration, including compatibility, completeness, and oversight. This review calls on the AI community to prioritize system integration as a core strategy for addressing AI risk, and to develop audit practices capable of capturing failures across components, environments, and systems beyond the reach of component-level evaluation.
2026-08-04
arXiv (preprint)
doi.org
arxiv.org