Publications

Don't Trust the Process: When Verifiability Undermines AI Accountability
Tamara Paris
AJung Moon
Jin Guo
How do we know if Artificial Intelligence (AI) systems are as performant and responsibly designed as the AI companies claim them to be? In a… (voir plus) race-driven innovation climate where responsive development requires time and resources, AI developers and providers may be tempted to misrepresent system performance or overstate their commitment to responsible AI principles. Such circumvention is further enabled by limited access to system components and information by external stakeholders, a restriction commonly justified on the grounds of trade secret protection, privacy and security considerations, among others. In response, a growing community of scholars has been developing cryptographic and statistical solutions that aim to enable robust verification of specific claims under constrained access. However, the construction of these solutions rely on a set of shared, yet unexamined, assumptions required to abstract complex real-world governance challenges into computational representations. In this article, we examine the validity of these assumptions. After detailing the conceptual foundation and analytical lens we used to interrogate these abstraction processes, we show that existing technical approaches to developing verifiable AI commit systematic fallacies that compromise the validity of these approaches. While the existing technical verification processes aim to solve critical AI governance problems, we argue that these fallacies create loopholes that can be exploited by dishonest developers and providers, and therefore lead to misplaced trust in these processes. Finally, we discuss how the field of verifiability could be reoriented towards a more nuanced and interdisciplinary approach to develop rigorous verification processes, both technical and non-technical, that support effective AI governance.
2026-06-24
ACM Conference on Fairness, Accountability, and Transparency (publié)
doi.org
Quantifying the Privacy of Counterfactuals by Leveraging Membership Inference Attacks Against Synthetic Data
Maryam Babaei
Yingke Wang
Hadrien Lautraite
Héber H. Arcolezi
Ulrich Aivodji
Sébastien Gambs
Counterfactuals are typically used in high-stakes decision areas to explain a machine learning model by showing how changes to the user prof… (voir plus)iles result in the desired outcome. However, explaining the model's decisions through counterfactuals can also be exploited by an adversary to conduct privacy attacks against the model or its training data. Drawing on the analogy that counterfactuals provide realistic substitutes for real training data, similar to synthetic data, we demonstrate in this paper how it is possible to successfully perform privacy attacks on counterfactuals by drawing on the attacks developed against synthetic data. More precisely, we investigate the effectiveness of the membership inference attacks designed for synthetic data on various types of counterfactuals. Additionally, while existing membership inference attacks against counterfactuals usually require to be able to query the model, we show how it is possible to perform successful membership inference attacks using only a set of counterfactuals, with no access to the model from which they are generated. Our results demonstrate that model developers should be more cautious when releasing counterfactuals to various users, as it can lead to a privacy breach.
2026-06-24
ACM Conference on Fairness, Accountability, and Transparency (publié)
doi.org
arxiv.org
Societal Frameworks Can Improve LLM Alignment
Karolina Stanczak
Nicholas Meade
Mehar Bhatia
Hattie Zhou
Konstantin Böttinger
Jeremy Barnes
Jason Stanley
Nicolas Papernot
Nicolas Chapados
Denis Therien
Timothy Lillicrap
Ana Marasovic
Sylvie Delacroix
Gillian K. Hadfield
Siva Reddy
Recent progress in large language models (LLMs) has focused on producing responses that meet human expectations and align with shared values… (voir plus) — a process coined alignment. However, aligning LLMs remains challenging due to the inherent disconnect between the complexity of human values and the narrow nature of the technological approaches designed to address them. Current alignment methods often lead to misspecified objectives, reflecting the broader issue of incomplete contracts, the impracticality of specifying a contract between a model developer and the model that accounts for every scenario in LLM alignment. In this paper, we argue that improving LLM alignment requires incorporating insights from societal alignment frameworks, including social, economic, and contractual alignment, and discuss potential solutions drawn from these domains. Given the role of uncertainty within societal alignment frameworks, we then investigate how it manifests in LLM alignment. It is this pervasive uncertainty that necessitates our alternative view on LLM alignment, framing the under-specified nature of its objectives as an opportunity rather than perfect their specification. Beyond technical improvements in LLM alignment, we discuss the need for participatory alignment interface designs.
2026-06-24
ACM Conference on Fairness, Accountability, and Transparency (publié)
doi.org
Tetradic Dynamics of Dyadic Sensorimotor Coordination: A Multiscale EEG Hyperscanning Study
Emmanuel Olarewaju
Lena Palaniyappan
Guillaume Dumas
How do leader and follower roles shape the brain mechanisms that support coordinated action between people? This question has direct therape… (voir plus)utic relevance for conditions such as schizophrenia and autism spectrum disorder, where the capacity for reciprocal social coordination is a defining vulnerability. Here, we propose a tetradic framework and examine sensorimotor coordination in 16 healthy adult pairs using simultaneous dual-brain EEG hyperscanning, a 2 x 2 within-subject design crossing Role (Leader/Follower) and Condition (Mirroring/Matching). Mirroring required resonance with a partner's movement; Matching required its controlled transformation. This contrast was designed to dissociate automatic from controlled coordination processes across roles. Behaviourally, Mirroring produced a reaction-time advantage that was selective to Followers, a finding replicated in a combined cohort, and consistent with role-dependent attention-inhibition gating. At the neural level, sustained (tonic) activity was dominated by Matching-related frontoparietal engagement regardless of role, while time-resolved (phasic) activity revealed a Mirroring-dominant reorganization that differentiated into role-specific patterns: anticipatory gating in Leaders and post-response inhibitory rebound in Followers. Information-theoretic decomposition of inter-brain coupling identified a Leader-specific predictive signal in medial prefrontal and cingulate cortices, a Follower-specific adaptive signal across sensorimotor and temporal regions, and a shared redundancy scaffold in orbitofrontal and insular cortices. These findings characterize tetradic coordination within dyads as a multiscale, role-asymmetric architecture in which top-down predictive control and bottom-up adaptive regulation are functionally dissociable. The tetradic framework provides an organizing scaffold for this dissociation, and the role-specific signatures it reveals offer candidate biomarkers for clinical populations in whom interpersonal coordination is disrupted.
2026-06-24
bioRxiv (prépublication)
doi.org
Data Selection Through Iterative Self-Filtering for Vision-Language Settings
Andrei Liviu Nicolicioiu
Sarvjeet Singh Ghotra
Morgane M. Moss
Aaron Courville
The availability of large amounts of clean data is paramount to training neural networks. However, at large scales, manual oversight is impr… (voir plus)actical, resulting in sizeable datasets that can be very noisy. Attempts to mitigate this obstacle to producing performant vision-language models have so far involved heuristics, curated reference datasets, and using pre-trained models. Here we propose a novel, bootstrapped method in which a CLIP model is trained on an evolving, self-selected dataset. This evolving dataset constitutes a balance of filtered, highly probable clean samples as well as diverse samples from the entire distribution. Our proposed Self-Filtering method iterates between training the model and selecting a subsequently improved data mixture. Training on vision-language datasets filtered by the proposed approach improves downstream performance without the need for additional data or pre-trained models.
2026-06-23
Transactions on Machine Learning Research (accepté)
doi.org
openreview.net
Lacuna: A Research Map for Machine Learning
Martin Weiss
Miles Q. Li
Alejandro H. Artiles
Yacine Mkhinini
Chris Pal
Hugo Larochelle
Nasim Rahaman
Lacuna is a research map for machine learning that uses LLMs to turn papers and scholarly metadata into markdown summaries, concept elements… (voir plus), research directions, and research proposals. Each item keeps links to the primary source records and papers that support it. We release the map with web, markdown, and MCP interfaces. Across LitSearch, Multi-XScience-CS/ML, and ScholarQA-CS-ML, Lacuna outperforms OpenScholar with the strongest gains on LitSearch retrieval (Recall@10 0.538 vs. 0.424 for OpenScholar v3). We also evaluate Lacuna Deep Research, a multi-stage report agent over the map, on 25 ReportBench-ML survey tasks: Lacuna Deep Research reaches 0.052 citation F1, 0.339 citation precision, 99 expert-reference hits, and 7.82/10 RACE report quality, while GPT-Researcher reaches 0.039 F1, 0.290 precision, 72 hits, and 5.24/10 RACE.
2026-06-23
arXiv (prépublication)
doi.org
arxiv.org
In situ cryogenic characterization of proton damage in thick p-channel skipper CCDs
Brandon M. Roach
Brenda Cervantes Vergara
Alex Drlica-Wagner
Phoenix Alpine
Ana Martina Botti
Claudio Chavez
Julian Cuevas-Zepeda
Juan Estrada
Guillermo Fernandez Moroni
Nora Hoch
Stephen E. Holland
Blas Irigoyen Gimenez
Agustín Lapi
Santiago Perez
Nathan Saffold
Javier Tiffenberg
Yikai Wu
Skipper charge-coupled devices (CCDs) are an offshoot of standard silicon pixel detectors and are capable of performing repeated non-destruc… (voir plus)tive charge measurements, enabling deeply sub-electron readout noise. This capability has opened the door to single-photon counting from the near-infrared (
2026-06-23
arXiv (prépublication)
doi.org
arxiv.org
AI <i>Automatons:</i> AI Systems Intended to Imitate Humans
Alicia DeVrio
Alexandra Olteanu
Solon Barocas
Su Lin Blodgett
Lisa Egede
Myra Cheng
2026-06-22
ACM Conference on Fairness, Accountability, and Transparency (publié)
doi.org
Con Moto: Embodied Steering of Music Transformers for Live Dance Improvisation
Zhixing Chen
Heidi Lei
Cheng-Zhi Anna Huang
Con Moto is a real-time generative music system for dance improvisation that supports embodied steering of a transformer model with configur… (voir plus)able levels of agency. While existing frameworks demonstrate the potential of embodied music-making and movement sonification in live performance, achieving both high musical coherence and low-latency responsiveness remains an ongoing challenge. In response, we leverage the musical coherence of real-time MIDI-based transformer models to design an integrated system that translates camera motion data into movement parameters, which in turn control the musical output. Con Moto employs two layers of control strategies: 1) inference-time steering of the transformer model and 2) post-generation rendering using Max/MSP as a control interface and Ableton Live for sound synthesis. We present the system through a duet performance for a live audience, supplemented by qualitative reflections from the dancers and the audience. By reconfiguring how the dancers' movements map to musical functions, we create a system with configurable agency. Fine-grained control over an individual musical voice invited dancers to experience the system as a playable instrument, while abstract musical mappings to a genre's energy opened space for the system to act as an autonomous creative partner. By navigating the aesthetic friction between human intent and AI agency, we explore a dynamic that facilitates a deep, bidirectional feedback loop.
2026-06-22
Zenodo (inconnu)
doi.org
Enhancing Expressive Musical Conversation in the jam_bot
Lancelot Blanchard
Perry Naseck
Katherine Liang
Joel Tan
Heidi Lei
Cheng-Zhi Anna Huang
Joseph Paradiso
Previous work introduced the jam_bot, a real-time system that embeds live music language models capable of generating symbolic music sequenc… (voir plus)es coherent with a performer's input. The system supports multiple interaction strategies that have been demonstrated in several public performances. However, these strategies limit expressive musical conversation by constraining tempo, form, or musical roles. We extend the jam_bot to support more expressive, open-ended interaction through four key improvements: (1) modeling velocity, a key dimension of expression in symbolic music; (2) increasing model throughput via a ggml implementation–required to accommodate the longer sequences induced by velocity modeling; (3) developing a new training modality that enables free-form call-and-response interaction across varying tempi; and (4) compensating for external MIDI output latency to ensure rhythmic coherence with the performer. We quantitatively evaluate the model throughput improvement and our latency compensation strategy, and offer MIDI samples online. Together, these enhancements enable the jam_bot to engage in natural, expressive musical conversation, eliminating key musical limitations to enable the development of future performances and installations.
2026-06-22
Zenodo (inconnu)
doi.org
A four-amino-acid YRIN insertion in ftsI reduced ceftiofur susceptibility in a canine Escherichia coli isolate
Xukun Dang
Zeshi Chen
Rina Bai
Junyao Jiang
Weishuai Zhai
Shizhen Ma
Lu Liu
Dejun Liu
Zhaofei Xia
Zhangqi Shen
Yang Wang
Siyu Chen
Abstract Ceftiofur resistance in canine clinical Escherichia coli is usually associated with extended-spectrum β-lactamases (ESBLs) or AmpC… (voir plus) β-lactamases. However, some isolates display elevated ceftiofur minimum inhibitory concentrations (MICs) without carrying these known resistance genes. In this study, we identified a phenotype-genotype discordant canine clinical E. coli isolate named 231255, with a ceftiofur MIC of 8 μg/mL. Routine resistance gene screening detected only bla TEM-1 and a chromosomal bla EC variant, bla EC-1149 , which could not adequately explain the elevated ceftiofur MIC. To investigate the underlying mechanism, a genomic library was constructed from genomic DNA of isolate 231255 and screened on ceftiofur-containing plates. Positive clones revealed one candidate determinant: an altered ftsI fragment encoding penicillin-binding protein 3 (PBP3) with a four-amino-acid YRIN insertion downstream of residue P333, previously identified in human E. coli . Previous studies have shown that this type of YRIN/YRIK insertion alone can reduce susceptibility to PBP3-targeting β-lactams, particularly aztreonam and ceftazidime. Functional validation showed that recombinant plasmids carrying the altered ftsI consistently increased the ceftiofur MIC to 4 μg/mL in different recipient backgrounds. These findings provide experimental evidence that ftsI /PBP3 alteration can elevate ceftiofur MIC in canine E. coli . Notably, the isolate belonged to ST410, and phylogenetic analysis indicated that it was not confined to a dog-associated background but instead clustered within a broader lineage shared across multiple sources, highlighting the need for potential dissemination of this mechanism and its associated resistant lineages at the human-companion animal interface.
2026-06-22
One Health Advances (publié)
doi.org
Discovering Latent Groups for Robust Classification
Ankur Garg
Ulrich Aivodji
Samira Ebrahimi Kahou
Vincent Michalski
Machine learning models exploit spurious correlations, achieving high average accuracy but failing disproportionately on underrepresented su… (voir plus)bgroups. Existing methods address this by adjusting network parameters, guided either by subgroup annotations or inferred pseudo-group labels. Yet at inference, these methods produce only a class prediction, with no insight into a sample's latent subgroup. We propose neural classification trees (NCT), a framework that achieves robustness by encoding subgroup structure in its tree-shaped architecture. By routing each sample to an "easy" or "hard" node of this tree -- based on prediction correctness -- and reusing these routes as pseudo-labels for the next iteration, NCT disentangles conflicting subgroups, without requiring subgroup supervision. We evaluate NCT on five benchmarks spanning binary and multi-class spurious correlations. Our experiments show that the learned tree topology provides strong interpretability by consistently isolating minority subgroups, which provides a transparent mapping between the model architecture and the data's latent group structure, while yielding competitive robustness with state-of-the-art methods.
2026-06-21
arXiv (prépublication)
doi.org
arxiv.org