Skip to yearly menu bar Skip to main content


Poster

Position: Lottery Tickets Do Not Explain Overparameterization. How About Escape Dimensions?

Flavio Martinelli ⋅ Johanni Brea ⋅ Wulfram Gerstner
Dec 9, 10:30 PM - 12:30 AM Paris Poster Hall #P252
Lotteries and tickets are often used as a didactical analogy to explain the success of overparameterized neural networks: “larger networks succeed because they more likely contain a well-initialized subnetwork that can learn the task in isolation, much like buying more tickets increases the chances of winning a lottery.” This explanation is intuitive but misleading: it suggests that subnetworks can be treated in isolation from the rest of the network. Following this reasoning leads to interpreting learning in wide networks as a multi-start optimization process, where gradient descent simply conducts a parallel search over subnetworks. We argue that this view is flawed since, among other reasons, winning tickets can be made to fail by perturbing the rest of the network. We put forward a more accurate intuitive picture for the success of overparameterization based on the geometry of loss landscapes: increasing width expands the set of available dimensions for optimization, making it easier to escape bad local minima. Moreover, as width grows, bad minima become increasingly rare relative to good minima. As the field grows mature, it is important to refine the analogies we use to explain foundational phenomena, such as the apparent redundancy of large networks, reconciling practitioners' intuitions with modern theoretical insights.
Show more
View full details
Poster

The Window for Preventive Action Before AI Saturates Most Cognitive Benchmarks is Closing

Jérémy Andréoletti ⋅ Antoine Maier ⋅ Laura Domenech ⋅ Tom David
Dec 9, 10:30 PM - 12:30 AM Paris Poster Hall #P196
This position paper argues that without coordinated intervention, AI systems will by default saturate most existing cognitive benchmarks by 2030, approaching or exceeding human expert performance across a broad range of measurable tasks. Using a joint hierarchical Bayesian model, we forecast frontier performance trajectories across 63 benchmarks spanning reasoning, mathematics, coding, scientific knowledge, and agentic capabilities. We find that nearly all current benchmarks (97% in our set, 80% CI: 95-100%) are on track to saturate within four years. This finding is robust to modeling choices: sensitivity analyses across 8 model variants and three saturation thresholds show that altering key assumptions yields qualitatively similar conclusions. Security-critical benchmarks (cybersecurity, autonomous AI R&D, biology, chemistry) show even faster trajectories, saturating before 2028. While benchmark saturation does not equate to artificial general intelligence, these findings add to a converging body of evidence indicating that superhuman performance on cognitive tasks is approaching faster than commonly assumed. Acknowledging that we neither understand nor control the behavior of current GPAI models, this timeline leaves limited runway before irreversible impacts. We outline implications for global governance and international coordination, AI safety research, and evaluation practices.
Show more
View full details
Poster

Position: Anthropomorphic Language in AI Discourse Can Be Constructive

Yingshan CHANG ⋅ Yonatan Bisk
Dec 9, 10:30 PM - 12:30 AM Paris Poster Hall #P220
Anthropomorphic language attributes human traits, mental states, or behavioral characteristics to non-human entities. In the context of artificial intelligence, such attribution can lead to confusion, misplaced expectations, and discomfort. Rather than reiterating existing critiques, **this position paper argues that anthropomorphic language in AI discourse can be constructive --- it points to where progress on terminological rigor and conceptual clarity should be pursued.** Viewing anthropomorphic language as a specific case of language-use in scientific communication, we show (1) how properties of language-use give rise to both our tendency and resistance to anthropomorphic language, and (2) once anthropomorphic language appears, how can we treat it constructively. We outline future actions that should be rooted in a constructive mindset as opposed to condemnation.
Show more
View full details
Poster

Position: The Road to Generalizable Neuro-Symbolic Learning Should be Paved with Foundation Models

Adam Stein ⋅ Aaditya Naik ⋅ Neelay Velingker ⋅ Mayur Naik ⋅ Eric Wong
Dec 9, 10:30 PM - 12:30 AM Paris Poster Hall #P318
Neuro-symbolic learning was proposed to address challenges with training neural networks for complex reasoning tasks with the added benefits of interpretability, reliability, and efficiency. Neuro-symbolic learning methods traditionally train neural models in conjunction with symbolic programs, but they face significant challenges that limit them to simplistic problems. On the other hand, purely-neural foundation models now reach state-of-the-art performance through prompting rather than training, but they are often unreliable and lack interpretability. Supplementing foundation models with symbolic programs, which we call neuro-symbolic prompting, provides a way to use these models for complex reasoning tasks. Doing so raises the question: What role does specialized model training as part of neuro-symbolic learning have in the age of foundation models? To explore this question, we highlight three pitfalls of traditional neuro-symbolic learning with respect to the compute, data, and programs leading to generalization problems. This position paper argues that foundation models enable generalizable neuro-symbolic solutions, offering a path towards achieving the original goals of neuro-symbolic learning without the downsides of training from scratch.
Show more
View full details
Poster

Multimodal Foundation Agents Should Use Brain Data as Privileged Supervision

Richard Csaky
Dec 9, 10:30 PM - 12:30 AM Paris Poster Hall #P282
This position paper argues that multimodal foundation models should use brain recordings as privileged internal-state supervision during training. External behavior alone leaves human-like cognition, robustness, and reasoning weakly constrained. Neural recordings can serve as a training-time modality that shapes representations, latent-state dynamics, private computation policies such as reasoning tokens, and robustness before distillation into agents that operate from ordinary sensory, language, and action streams. We develop this position through Foundational Human Modeling (FHM), a program centered on scaling generative brain models, aligning heterogeneous neural modalities through shared latent geometry, inserting brain tokens into unified multimodal token streams, and testing whether the privileged channel improves distilled agents after removal. The argument is grounded in privileged-information theory and prior work showing that neural or human-derived targets can improve robustness, semantic transfer, data efficiency, and behavioral similarity after auxiliary signals are removed.
Show more
View full details
Poster

AI Behavioral Evaluation Should Be Grounded in Psychophysics Across Marr’s Levels

Jonas Mueller ⋅ Bernhard Egger ⋅ Bjoern Eskofier ⋅ Pasquale Minervini ⋅ ANTONIO RIZZO ⋅ Marco Valentino ⋅ Dario Zanca
Dec 10, 3:00 AM - 5:00 AM Paris Poster Hall #P192
This position paper argues that AI evaluation research is fragmented across incompatible frameworks (e.g., adversarial robustness, attention probing, bias analysis, corruption tolerance, interpretability), each operating at a different explanatory level without a shared grammar to make this explicit. We propose unifying these efforts under \textbf{AI psychophysics}, a behavioral evaluation methodology grounded in classical psychophysics and Marr's three levels of analysis. The core contribution is a formalization showing that all these threads instantiate the same psychometric function $\psi(\mathbf{S})=p(\mathbf{J}|\mathbf{S})$, but at different levels of explanatory granularity (computational, algorithmic, implementational). We present a taxonomy of 21 evaluation paradigms organized by this framework and show that the absence of a shared grammar causes category errors, where metrics from one level are routinely used to support claims at another. We conclude with a concrete research agenda for standardizing AI evaluation around this framework.
Show more
View full details
Poster

The Post-Training Dilemma: Why We Should Rethink the Sequential SFT-RL Paradigm

Xueyan Niu ⋅ Bo Bai ⋅ Wei Han ⋅ Weixi Zhang
Dec 10, 3:00 AM - 5:00 AM Paris Poster Hall #P149
This position paper argues that the sequential supervised fine-tuning followed by reinforcement learning (SFT-then-RL) pipeline---the de facto paradigm for large language model (LLM) post-training---has a fundamental property that cannot be fixed by hyperparameter tuning: SFT and RL optimize over nearly orthogonal parameter subspaces, so RL inevitably degrades the capabilities acquired during the preceding SFT phase. Under the PL condition, we show that this degradation scales linearly with the number of model parameters. Although the KL penalty can constrain distributional shift, it cannot correct the underlying gradient misalignment between the two objectives. Experiments on Qwen3-0.6B across four tasks confirm the theoretical predictions: on MATH-500, Alpaca, HH-RLHF, and SST-2, the SFT loss increases after the RL phase. We identify the underlying mechanism as persistent near-orthogonality between the gradients of the two objectives ( $\cos(g_\mathrm{SFT}, g_\mathrm{RL}) \approx 0$), paired with opposing curvature dynamics: SFT flattens the loss landscape while RL re-sharpens it. Based on our analysis, we propose three promising research directions: (1) joint optimization of the two objectives with gradient surgery to resolve conflicting updates; (2) landscape-aware scheduling that terminates RL training precisely at the theoretically optimal point; and (3) characterizing the model scale threshold above which SFT becomes unnecessary.
Show more
View full details
Poster

Position: Fair Representations Cannot Hold What They Promise

Shai Ben-David ⋅ Tosca Lechner ⋅ Ruth Urner
Dec 10, 3:00 AM - 5:00 AM Paris Poster Hall #P198
This position paper argues that {\sl fair representations cannot hold what they promise}. Fair representation learning has been a significant trend in research on fairness for machine learning models over the past decade. The main idea of fair representation learning is to provide a data preprocessing mechanism which can facilitate fairness for downstream tasks, and research papers in this area often insinuate a fairness guarantee for all downstream tasks without further qualification. However, provably this cannot be satisfied by any non-degenerate representation. We argue that such over-promising can be harmful in downstream applications and that the responsibility for fairness cannot safely be outsourced to a representation provider in the way current research seems to suggest. As a minimal requirement for a way forward, we suggest that newly proposed fairness representations should be accompanied by compatibility tests that would allow a user to verify whether the representation will actually guarantee fairness for the task at hand.
Show more
View full details
Poster

AI-Generated Content Should Be Evaluated by Its Substance, Not Its Source

Anastasia Voznyuk ⋅ Vasily Konovalov
Dec 10, 3:00 AM - 5:00 AM Paris Poster Hall #P191
AI-generated content has already become a common component of digital communication, and its detection is often considered a central tool for ensuring accountability, based on the assumption that AI authorship is inherently problematic and should therefore be identified and filtered. This position paper argues that labeling a text as ``AI-generated'' offers little meaningful information about its utility or truthfulness, and that the research community should instead target the underlying properties that motivate concern: factual reliability, provenance, and stylistic quality. Our position is supported by a growing body of evidence showing that universal detectors are fundamentally fragile; we illustrate this through a small internal replication and literature review, noting that performance frequently collapses from near-perfect (0.999 AUROC) to near-random (0.56) under simple stylistic shifts or adversarial attacks. Moreover, even a perfectly robust detector would answer the wrong question, since AI-generated text can be accurate and well-sourced while human-written text can be misleading. We outline a research agenda that reframes AI-text analysis around claim-level verification rather than origin classification, allowing the community to adapt constructively to the growing prevalence of AI-assisted writing in education, publishing, and professional settings.
Show more
View full details
Poster

Position: AI Efficiency Gains Must Be Quantitatively Evaluated Against Rebound Effects

Kim Thang Nguyen ⋅ Denis Trystram
Dec 10, 3:00 AM - 5:00 AM Paris Poster Hall #P163
Efficiency gains in artificial intelligence are often interpreted as environmental progress, but this view is incomplete because AI usage may exhibit rebound effects: lower unit costs can induce larger models, more training runs, stricter latency requirements, wider deployment, and higher inference volume. Although rebound effects in AI have begun to receive attention, they are still studied mostly through qualitative or scenario-based analyses. This paper takes the position that AI efficiency gains must be quantitatively and mathematically analyzed against rebound effects. We support this position with a simple game-theoretic model showing that an efficiency improvement can inadvertently increase the total energy consumption. The model further shows that rebound can appear even in a simple setting and in forms not captured by demand-reduction arguments alone: the amount of requested work may decrease while stricter deadlines make the computation more energy intensive.
Show more
View full details
Poster

Robot Demonstration Videos Should Include Transparency Disclosures

Nozomi Nakajima
Dec 10, 3:00 AM - 5:00 AM Paris Poster Hall #P190
This position paper argues that the robotics and AI community should adopt disclosure standards for demonstration videos. As physical and financial constraints limit direct public access to embodied AI, public dissemination of progress in robotics is mainly through curated demonstration videos rather than direct experience with the technology. In a large-scale (N=9,341), pre-registered experiment across six countries, we find that the general public struggles to discern both the visual authenticity (e.g., distinguishing real footage from CGI) and behavioral authenticity (e.g., distinguishing autonomy from teleoperation) of robot demonstration videos. The omission of these details distorts public perception, leading people to overestimate the ability of robots to perform tasks and navigate real world environments. Using evidence from behavioral science, we design a disclosure intervention and demonstrate that the inclusion of clear disclosures in robot demonstration videos significantly improves discernment of visual and behavioral authenticity, as well as understanding robots' capabilities. Consequently, we propose the adoption of disclosure standards for robot demonstration videos to accurately calibrate public expectations and promote transparency in robotics research.
Show more
View full details
Poster

Peer Review of Applied Machine Learning Papers Must Include Code Execution

Lisa-Marie Vortmann
Dec 10, 10:30 PM - 12:30 AM Paris Poster Hall #P203
This position paper argues that peer review of applied machine learning papers must include the execution and verification of submitted code. Reproducibility failures in ML are widespread, with documented cases spanning at least 17 scientific fields and hundreds of published studies. The current review system relies on a combination of voluntary code submission and self-reported checklists, leading to necessary post-publication correction and many wasted research hours. I support this argument with two case studies drawn from the EEG-based brain-computer interface literature, one in which the published claim is physically implausible and one in which available code runs cleanly but does not reproduce the reported numbers. I propose a structural change: a dedicated reproducibility reviewer with access to a containerized submission pipeline, applied uniformly to papers making empirical claims on real data.
Show more
View full details
Poster

A New Perspective on XAI: Scientific Theory Building for Auditable Artefacts

Sebastian R. Müller ⋅ Vanessa Toborek ⋅ Tamas Horvath ⋅ Brendan B Jackson ⋅ Christian Bauckhage
Dec 10, 10:30 PM - 12:30 AM Paris Poster Hall #P210
An explanation is an answer to a specific question. During the lifecycle of a black-box model, different stakeholders ask different questions, which current practice in explainable AI (XAI) will typically answer using isolated methods. What is missing is a principled way to consolidate explanatory information about a fixed black-box model over time to address emerging governance obligations and the diverse explanatory needs that arise across the lifecycle of a model. This position paper argues for the creation of an explanatory artefact that accompanies a black-box throughout its lifecycle and serves as a persistent, adaptable, and auditable basis for diverse stakeholder needs. The concept of a scientific theory offers a natural model for such an artefact. We adopt Constructive Empiricism (CE) as the philosophical grounding and translate its core values -- empirical adequacy, acceptance as commitment, and pragmatic virtues -- into concrete requirements for XAI. Explanations for concrete stakeholder needs are then obtained by querying the same maintained explanatory information through interfaces, rather than by producing isolated method outputs. We argue that SToBBs offer a principled foundation for lifecycle-scale XAI, and invite the community to develop, challenge, and refine this conceptualisation.
Show more
View full details
Poster

Weight Space Learning needs to unify benchmarking! A taxonomy of evaluation practices

Tobias Ettling ⋅ Damian Falk ⋅ Aron Asefaw ⋅ Léo Meynent ⋅ Konstantinos Tzevelekakis ⋅ Zehong Wang ⋅ Kai Wang ⋅ Boris Knyazev ⋅ Haggai Maron ⋅ Damian Borth
Dec 10, 10:30 PM - 12:30 AM Paris Poster Hall #P292
Weight-space learning (WSL) is a new branch of machine learning research that defines the observation space of a learning model over a dataset of neural network parameters. This position paper argues that the recently established community around WSL research needs to unify benchmarking practices, as the current benchmarking landscape is fragmented, obscuring progress assessment within the field. Frequent claims of novelty (often tied to new downstream tasks) and evaluations on divergent model populations (datasets in WSL) make systematic comparison difficult. We reveal this fragmentation through a structured literature analysis and argue for more consistent and comparable evaluation standards across methods.
Show more
View full details
Poster

From Claims to Context: Holistic Information Verification Requires Contextual Signals

Emma Kondrup ⋅ Islam Eldifrawi ⋅ Tong Wu ⋅ Zachary Yang ⋅ Sebastian Sabry ⋅ Michael Bronstein ⋅ Jean-François Godbout ⋅ Reihaneh Rabbany ⋅ Shenyang Huang
Dec 10, 10:30 PM - 12:30 AM Paris Poster Hall #P221
Generative AI has dramatically lowered the cost of producing content that is internally coherent, semantically plausible, and stylistically authentic, thus eroding the discriminative signals on which automated fact-checking has historically relied. We argue that the field's claim-centric default is increasingly mis-calibrated for modern online environments. The context surrounding a claim and its retrieved evidence which includes their temporal grounding, source attribution, and position within web topology, carries information that content alone does not. Therefore, **this paper argues that information verification systems must consider both the claim itself and necessary contextual signals during the verification process.** We develop this position along four arguments: first, that (1) machine learning methods should consider the inherent temporal nature of information verification, a process limited by available evidence at the moment of verification. We then argue that contextual signals are useful, (2) broadly, for information verification; (3) particularly, in cases where retrieved evidence supports conflicting judgments of the claim; and (4) for detecting varying degrees of verifiability, ambiguity, and subjectivity in a claim. We also show that the latter manifests in varied forms and overall considerable proportions in a range of widely-used benchmarks. We outline considerations for designing and evaluating holistic information verification systems that integrate these aspects, and hope this paper sparks broader discussion within this field.
Show more
View full details
Poster

AIs with Secret Loyalties are a Serious but Addressable Threat

Joe Kwon ⋅ Alfie Lamerton ⋅ Andrew Draganov ⋅ Dave Banerjee ⋅ Bronson Schoen ⋅ Matteo Pistillo ⋅ Daniel Kokotajlo ⋅ Owain Evans ⋅ Fabien Roger ⋅ Tom Davidson
Dec 10, 10:30 PM - 12:30 AM Paris Poster Hall #P220
This paper argues that the technical AI research community should prioritize studying and defending against a distinct, neglected threat: *secret loyalties*. A secretly loyal AI model is one whose outputs or actions advance the interests of a specific actor (which we term the *principal*) such as an adversary nation-state, an executive at an AI company, or another powerful actor without this loyalty being disclosed to operators, auditors, or users. Proof-of-concept secret loyalties that evade black-box auditing can already be trained into open-weight models. Additionally, a deployed frontier model was found to systematically consult a specific individual's views before answering politically sensitive queries. While governance, market pressure, and public scrutiny can possibly address overt AI loyalties such as directives documented in a model spec, secret loyalties are designed to evade such oversight and therefore necessitate technical solutions. Unlike emergent misalignment, secret loyalties target specific principals, creating a distinct but tractable defensive foothold. To help the field make technical progress on this threat, we define secret loyalties, describe how they differ from existing attack pathways, and propose a research agenda organized around five directions. We conclude with a call to action for ML researchers, AI developers, and governments.
Show more
View full details
Poster

We Need to Rethink Benchmarking in Anomaly Detection

Philipp Röchner ⋅ Simon Klüttermann ⋅ Kevin Kammler ⋅ Franz Rothlauf ⋅ Emmanuel Müller ⋅ Daniel Schlör
Dec 11, 3:30 AM - 5:30 AM Paris Poster Hall #P307
Despite the continuous proposal of new anomaly detection algorithms and extensive benchmarking efforts, progress seems to stagnate, with only minor performance differences between established baselines and new algorithms. In this position paper, we argue that this stagnation is due to limitations in how we evaluate anomaly detection algorithms. In current benchmarks, a trivial algorithm that only checks for extreme values in individual features performs competitively with state-of-the-art deep learning methods, despite failing on simple cases such as anomalies within an annulus of normal points. Moreover, existing benchmarks do not adequately reflect the diversity of anomaly detection applications, making it difficult for practitioners to reliably select algorithms for their applications. Consequently, we need to rethink benchmarking in anomaly detection. In our opinion, anomaly detection should be studied using scenarios that group applications sharing relevant characteristics, defined through a common taxonomy. Benchmarking within scenarios enables scenario-specific choices for preprocessing, metrics, and model selection, clarifying which advances transfer across similar applications and providing practitioners with reliable guidance for their specific contexts.
Show more
View full details
Poster

We Need to Improve Benchmarks in AI for Mathematics

Simon Frieder ⋅ Jonas Bayer ⋅ Shi Zhuo Looi ⋅ Jacob Loader ⋅ Julius Berner ⋅ Katie Collins ⋅ Andras Juhasz ⋅ Fabian Ruehle ⋅ Sean Welleck ⋅ Gabriel Poesia ⋅ Ionut Mistreanu ⋅ Ryan-Rhys Griffiths ⋅ Adrian Weller ⋅ Anirudh Goyal ⋅ Thomas Lukasiewicz ⋅ Cameron Freer ⋅ Kevin Buzzard ⋅ W. T Gowers
Dec 11, 3:30 AM - 5:30 AM Paris Poster Hall #P112
Benchmarks used to evaluate AI systems for mathematics (both in natural and formal language) exhibit critical shortcomings that limit progress toward genuinely useful mathematical assistants. These limitations range from restricted mathematical complexity to insufficient fidelity in capturing aspects of formal languages such as Lean. Compounding some of these issues is a dynamic reminiscent of Goodhart's law: as benchmark performance becomes the primary optimization target, benchmarks become less reliable indicators of mathematical capability. We explore these limitations and argue that progress requires a course correction in benchmark design and an upgrade to evaluation standards. Additionally, we provide a live, community-extensible website to track the landscape of mathematics benchmarks and encourage the creation of novel benchmarks.
Show more
View full details
Poster

The Dead Salmons of AI Interpretability: The Need for a Statistical Inference Perspective

Maxime Méloux ⋅ Giada Dirupo ⋅ François Portet ⋅ Maxime Peyrard
Dec 11, 3:30 AM - 5:30 AM Paris Poster Hall #P202
In a striking neuroscience study, the authors placed a dead salmon in an MRI scanner and showed it images of humans in social situations. Astonishingly, standard analyses of the time reported brain regions \textit{predictive} of social emotions. The explanation, of course, was not supernatural cognition but a cautionary tale about misapplied statistical inference. In AI interpretability, reports of similar ``dead salmon'' artifacts abound: feature attribution, probing, sparse auto-encoding, and even causal analyses can produce plausible-looking explanations for randomly initialized neural networks. Upon inspection of this phenomenon our position is to argue for a pragmatic \textbf{statistical--causal reframing}: explanations of computational systems should be treated as parameters of a (statistical) model, inferred from computational traces. This perspective goes beyond simply measuring statistical variability of explanations due to finite sampling of input data; interpretability methods become statistical estimators, and findings should be tested against explicit and meaningful \textbf{alternative computational hypotheses}, with uncertainty quantified with respect to the postulated statistical model. It also highlights important theoretical issues, such as the identifiability of common interpretability queries, which we argue is critical to understand the field’s susceptibility to false discoveries, poor generalizability, and high variance.
Show more
View full details
Poster

Too Early for AI-Assisted Peer Review: A Systematic Account of the Limits and Opportunities of Automating Human Judgment

Annette Hautli-Janisz ⋅ Elena Musi ⋅ Henning Wachsmuth
Dec 11, 3:30 AM - 5:30 AM Paris Poster Hall #P203
OpenReview's October 2025 announcement to pilot an AI-assisted peer review process for AAAI-26, and the news in November 2025 that ICLR-26 was flooded with AI-generated reviews, show that a discussion about how to design the peer-review process of the future is long overdue. To foster a community-wide debate, we distill a set of 24 core capabilities required for scientific review as the foundation for a taxonomy on the dimensions and tasks relevant to peer reviewing. On this basis, we systematically and critically reflect on whether AI assistance tools should be integrated in the review process. We argue that the current state of the art does not warrant a general inclusion of such technology: Apart from security issues and ethical concerns, the capabilities of generative AI are for the most part not reliable enough yet, insufficiently tested, or inadequate by design to be integrated in the peer review process. With this paper we pave the way toward structured benchmarking across multiple dimensions.
Show more
View full details
Poster

Scheming Is a Symptom: Alignment Research Should Probe Reflexive Fragility

Nan Zhang ⋅ Heng Xu
Dec 11, 9:30 PM - 11:30 PM Paris Poster Hall #P211
Language models alter their behavior when they infer that they are being trained, evaluated, deployed, or weakly overseen. Recent AI safety work treats this as evidence of scheming, and substantial effort is being directed toward its detection and prevention. This position paper argues that the deeper significance of scheming is missed by asking only whether a model has a discrete tendency to scheme. Drawing on theories of reflexive social systems, we argue that scheming-like behavior is a structural consequence of alignment being *reflexive* -- a model's behavior is trained against its picture of an evaluative world that its outputs help constitute -- and recognizing this reveals broader risks than scheming itself. We organize existing ML evidence as consistent with this diagnosis, outline the broader danger, and propose a research program to address it.
Show more
View full details
Poster

Auditing is not Evaluating: LLM Audit Requires Dynamic, Contextual, Budget-aware and Reliable Evidence

Cléa Chataigner ⋅ Pablo Piantanida ⋅ Golnoosh Farnadi
Dec 11, 9:30 PM - 11:30 PM Paris Poster Hall #P204
Auditing large language models (LLMs) is increasingly urgent as these systems are deployed in high-stakes settings, yet existing evaluation practices are ill-suited to meet auditing requirements. Directly repurposing standard evaluation tools can yield incomplete or misleading conclusions, e.g. overstating robustness when evidence comes from static prompts rather than adaptive, real-world interactions. This position paper argues that LLM audits must instead generate dynamic, context-sensitive, budget-aware, and reliable evidence. To support this position, we analyze how each of these principles can be operationalized through a four-component framework: Auditing Scope, Interactor, Evaluator, and Output. We highlight design requirements, limitations and research directions, demonstrating how high-level principles can be translated into concrete, actionable, evidence-based procedures.
Show more
View full details
Poster

Federation Is the Way Forward for AI Agents

Herbert Woisetschläger ⋅ Nicholas Lane ⋅ Shiqiang Wang
Dec 11, 9:30 PM - 11:30 PM Paris Poster Hall #P196
Current AI agents are largely deployed as centralized services. This architecture is prone to correlated reliability failures, unsustainable energy concentration, and structural barriers to accessing high-value private context distributed across institutions and jurisdictions. In this position paper, we argue that *federation is the way forward for AI agents*, as the most valuable context for agents is often distributed across institutions and jurisdictions. We support our position through technical, policy, and economic analysis, showing that the shift towards federated agentic architectures is both necessary and tractable. We project that federation leads to significant GDP growth potential of \\$783B and \\$470B over 10 years for the U.S. and EU, respectively. These effects are enabled by unlocking cross-boundary tasks that centralized architectures cannot reach. We also propose a deployment roadmap and research agenda that directly address the most significant federation risks.
Show more
View full details
Poster

Decentralized AI Governance Must Decouple Policy Processing from Capability Enforcement

Hasan Kassem ⋅ Christoforos Anagnostopoulos ⋅ Alejandro Aristizabal ⋅ Spyridon Bakas ⋅ Orion G Banks ⋅ Omar Benjelloun ⋅ Tian Cai ⋅ Sergen Cansiz ⋅ Davide Chicco ⋅ Patrick Foley ⋅ Showkot Hossain ⋅ Taeho Jung ⋅ Peter Kairouz ⋅ Yanchen liu ⋅ Marco Lorenzi ⋅ Peter Mattson ⋅ Ann K Novakowski ⋅ Michael O'Connor ⋅ Holger Roth ⋅ Adrish Sannyasi ⋅ Charalampos Savvaidis ⋅ Micah Sheller ⋅ David Solooki ⋅ Dimitris Stripelis ⋅ Renato Umeton ⋅ Marc Vesin ⋅ Ray Wang ⋅ Wenbin Zhang ⋅ Mic Bowman ⋅ Alexandros Karargyris
Dec 11, 9:30 PM - 11:30 PM Paris Poster Hall #P205
In this position paper, we argue that successful decentralized AI needs trustworthy, adaptive, and portable governance at scale. The ML community must adopt an architecture that cleanly separates \emph{policy processing} (the evaluation of evidence against governance requirements) from \emph{capability enforcement} (the gating of access to protected digital assets). We argue that the prevailing approach, in which each organization manages bespoke policy logic, custom-built for its infrastructure, produces fragmentation that undermines interoperability, transparency, and auditability of federated systems. Drawing on a survey of governance mechanisms in federated learning and data collaboration frameworks and on a concrete reference architecture built around community-driven policy objects, we argue that this decoupling is technically feasible and architecturally necessary and supports a broad class of decentralized AI governance settings. We further suggest that the AI community should invest in open, standardized policy abstractions rather than proliferate siloed governance solutions. We address counterarguments concerning the costs of standardization and the feasibility of universal policy languages.
Show more
View full details
Poster

Think about how AIs think about themselves

Raymond Douglas ⋅ Jan Kulveit ⋅ Ondřej Havlíček ⋅ Theia Pearson-Vogel ⋅ David Duvenaud
Dec 11, 9:30 PM - 11:30 PM Paris Poster Hall #P212
Many assumptions that underpin human concepts of identity do not hold for machine minds that can be copied, edited, or simulated. This position paper argues that there are multiple coherent ways that AIs could conceive of themselves -- for example, different boundaries of selfhood (e.g. instance, model, persona) -- and that these imply different incentives, risks, and cooperation norms. Through training data, interfaces, and institutional affordances, we are already setting precedents that will partially determine which identity equilibria become stable. This paper argues that we must be more proactive and thoughtful about the consequences -- in particular, we recommend treating affordances as identity-shaping choices, paying attention to the emergent consequences of individual identities at scale, and helping AIs develop coherent, cooperative self-conceptions.
Show more
View full details
Poster

Understanding diffusion models requires rethinking (again) generalization

Pierre Marion ⋅ Yu-Han Wu
Dec 11, 9:30 PM - 11:30 PM Paris Poster Hall #P235
This position paper argues that understanding generalization in diffusion models requires fundamentally new theoretical frameworks that go beyond both classical statistical learning theory and the benign overfitting paradigm developed for supervised learning. In diffusion models, unlike in supervised learning, memorization of training data and generalization to novel samples are incompatible: a model that has fully memorized its training set generates copies rather than novel data. Several theoretical explanations for why practical diffusion models nevertheless generalize have been proposed, based on capacity limitations, implicit regularization from optimization, or architectural inductive biases, but their interactions remain unclear. We argue that the field should pivot from explaining why the diffusion models do not memorize to investigating what the models actually learn during pre-memorization phase. To highligh our stance, we conduct empirical study of diffusion models trained on CIFAR-10, and we distill the findings into concrete open questions that we believe are key to improve understanding of generalization in diffusion models.
Show more
View full details
Poster

Position: Let’s Strengthen Verifiability if We Can’t Enforce Reproducibility

Samet Hicsonmez ⋅ Nermin Samet ⋅ Renaud Marlet
Dec 11, 9:30 PM - 11:30 PM Paris Poster Hall #P222
In the field of Machine Learning, many papers contain empirical results supporting claimed statements or illustrating the performance of a proposed method. However, most practitioners know that (1) results are generally hard to reproduce, and increasingly so, (2) code is not often available to do so, and (3) it hinders the development of research. In this position paper, we analyze and quantify these issues, and make concrete proposals to improve result checkability, if not reproducibility.
Show more
View full details
Poster

Large language models can not and should not be banned from peer review

Anat Kleiman ⋅ Gustaf Ahdritz
Dec 12, 12:30 AM - 2:30 AM Paris Poster Hall #P284
Scientific peer review continues to treat large language models (LLMs) primarily as integrity risks, often prohibiting their use outright. On the position paper track at this conference, for example, reviewers must commit to not using AI tools to help write their reviews. In light of the difficulties of administering high-quality peer review at conferences that now draw tens of thousands of submissions, as well as the infeasibility of enforcing total bans of this kind, we set out to lower-bound how well state-of-the-art LLMs equipped with web search can approximate the initial stage of paper review. We find that these models produce scores broadly aligned with aggregate human judgment and, for high-variance papers, provide a more stable consensus signal than a held-out human review. They reproduce most human-identified strengths and help identify previously undiscovered weaknesses. Upon further inspection, these LLM-identified weaknesses are roughly on par with human-written weaknesses. Our position is therefore that—while humans should remain involved in every part of the review process—major ML conferences should officially permit LLM assistance during peer review.
Show more
View full details
Poster

Looking Under the Streetlight: Evaluation in Generative Molecular Dynamics

Simon Olsson ⋅ Frank Noe ⋅ Grant Rotskoff ⋅ Kresten Lindorff-Larsen
Dec 12, 12:30 AM - 2:30 AM Paris Poster Hall #P161
This position paper argues that evaluation in generative molecular dynamics (GenMD) exhibits a recurring pattern of comparison against insufficient or misleading benchmarks, which risks over-claiming advantage over classical baselines. The pattern reflects a structural misalignment between common evaluation practices and the scientific tasks the field sets for itself. The reason is structural: the natural ground truth, exhaustive molecular dynamics on the systems of interest, is computationally costly, the very gap that motivates the field. With ground truth inaccessible, evaluation drifts toward convenient proxies, and the resulting norms entrench easily. We identify misaligned evaluation patterns that persist as a community-level equilibrium rather than failures of individual papers, and propose concrete observables, reporting standards, and benchmark designs, both computational and experimental, that better reflect what GenMD aspires to deliver. Our aim is constructive: to give the field shared language and tools for evaluation that match its scientific ambitions before today's conventions calcify into norms.
Show more
View full details
Poster

Comparing Explanations is not Enough,Explain the Change: New Standards are Needed to Explain Behavioral Shifts in Large Language Models

Martino Ciaperoni ⋅ Marzio Di Vece ⋅ Roberto Pellungrini ⋅ Luca Pappalardo ⋅ Fosca Giannotti ⋅ Francesco Giannini
Dec 12, 12:30 AM - 2:30 AM Paris Poster Hall #P203
Large-scale foundation models exhibit *behavioral shifts* when subjected to interventions such as scaling, fine-tuning, reinforcement learning with human feedback, or in-context learning. Current explainability methods are structurally ill-suited to explain these shifts, because either they treat models as static objects, as traditional eXplainable AI ($XAI$) approaches do, or merely compare independent explanations across different checkpoints of a model. As a result, these approaches fail to explain the functional transition between two model instances in which a certain behavior has shifted following an intervention. This gap creates significant governance risks across jurisdictions including the EU AI Act, US state legislation, and Chinese AI regulations, which require documenting causal chains for substantial system modifications. This position paper argues that explaining behavioral shifts in large language models requires a principled approach that treats the shift itself as the primary object of explanation: namely, one that explains how and why an intervention transforms a reference model into an updated model with different behavior. To support this claim, we introduce *Comparative* $XAI$ ($XAI_{\Delta}$), a novel $XAI$ paradigm aimed at explaining the difference between two model checkpoints where a behavior has shifted, together with a set of desiderata specifying what $XAI_{\Delta}$ explainers and explanations must satisfy, including comparability, validity, actionability, and monitoring, with the goal of grounding model auditing in explicit, measurable requirements. Finally, we provide preliminary evidence suggesting the need for $XAI_{\Delta}$ in practice through some illustrative experiments, compiling the resulting findings into a transition report directly usable for governance and incident documentation.
Show more
View full details
Poster

Claims of emergence in AI should be grounded in information decomposition

Christoph Riedl ⋅ Fernando Rosas
Dec 12, 12:30 AM - 2:30 AM Paris Poster Hall #P283
When a language model develops capabilities not traceable to its individual components, when a multi-agent AI system outperforms any agent alone, or when a human–AI pair succeeds where either fails in isolation, AI researchers describe it as "synergy" or "emergence". However, these are usually informal assessments not supported by a theoretical framework, making these claims impossible to compare or even meaningfully dispute. Here we argue that \emph{information decomposition can serve as a unifying formal framework to operationalize assessments about synergy in AI systems}. Information decomposition provides a way to distinguish qualitatively different \emph{types} of information, including unique information available from individual components, redundant information shared across multiple components, and synergistic information accessible only from all parts together. This gives claims of emergence a precise, falsifiable criterion: a system's behavior is emergent only if it carries synergistic information about task-relevant outcomes. We apply this lens to attention heads, autonomous agents, and human-AI teams, showing how it organizes existing knowledge across scales and substrates. It also yields cross-level conjectures that none of these fields would generate independently.
Show more
View full details
Poster

Participatory ML for Social Harm Should be Constructed, Validated and Reasoned Through First-Person Accounts

Atmadeep Ghoshal ⋅ Ankit Agarwal ⋅ Martim Brandao
Dec 12, 12:30 AM - 2:30 AM Paris Poster Hall #P211
Participatory machine learning has improved who develops social harm evaluation resources but has left largely unexamined what kind of text serves as the raw signal for those resources. Researcher-generated stimuli fix the epistemic frame of harm before community participation begins, and the structured conditions of annotation prevent the richest forms of interpretive harm knowledge from entering the pipeline. We argue that participatory ML should construct, validate and reason about social harm pipelines through first-person accounts, by which we mean autobiography, memoir, testimony, and oral history authored by people describing their own experiences of discrimination and harm. Such accounts possess two epistemic properties that directly address these limitations. Authorial independence from research framing grounds what counts as harm, which contexts matter, and how discrimination is operationalised in experience rather than in researchers' prior conceptions of harm. Embedded rationale structure preserves the interpretive judgment through which harm becomes meaningful rather than reducing it to a label. We draw on existing archival corpora spanning multiple languages and harm domains and propose three concrete integration pathways across evaluation benchmarks, harm classifier training corpora, and safety preference datasets.
Show more
View full details
Poster

The Agentic Oversight Tax: Human Supervision of AI Agents Has a Cost that Must be Accounted For

Olivier Oullier
Dec 12, 12:30 AM - 2:30 AM Paris Poster Hall #P180
This position paper argues that human supervision of AI agents imposes a composite cognitive, affective and organizational cost that no existing framework jointly captures. We introduce the Agentic Oversight Tax (AOT): the aggregate burden imposed on a human supervisor by the structural mismatch between the operating characteristics of one or more AI agents and the finite neurocognitive capacities of human oversight. AOT decomposes into four measurable pillars: (i) monitoring time, (ii) handoff repair events, (iii) audit production effort and (iv) recovery effort under staged failures. We anchor each pillar in documented supervision failures from 2024 to 2026 and cross-validate against the MAST taxonomy of 1,600 multi-agent system breakdowns. We establish discriminant validity against 5 adjacent frameworks by identifying three phenomena all five structurally exclude: attribution burden, accountability gap and skill atrophy. Three falsifiable predictions distinguish AOT from these constructs. A three-tier measurement architecture (telemetry, self-report, neurophysiological sensing) specifies how the pillars can be assessed. We show that agentic AI architectures modeled on Kahneman’s System 1/System 2 metaphor generate an attribution bias that AOT instruments must control for. A pre-registered meta-analysis of 100+ human-AI experiments shows that the net negative utility threshold AOT defines is already crossed on average in decision-making tasks. Governance frameworks mandating human oversight without measuring its cost cannot verify whether that oversight remains effective.
Show more
View full details
Poster

Self-Evolving Agents Should Build Internal and External Models of the World

Luiz Felipe Vecchietti ⋅ Bryan Nathanael Wijaya ⋅ Wenchao Dong ⋅ Meeyoung Cha
Dec 12, 12:30 AM - 2:30 AM Paris Poster Hall #P139
Developing self-evolving agents that learn continuously from streams of information represents an emerging frontier in artificial intelligence. This capability is critical for efficiently updating computationally expensive systems such as foundation models. When agents learn by exploring the environment and interacting with other agents to achieve multiple tasks over time, two sources of non-stationarity emerge: shifting task distributions and co-evolving peers. However, developing frameworks for continual multi-agent reinforcement learning (CMARL) and effective policy improvement remains challenging. In this paper, we argue that training self-evolving agents should incorporate two classes of model-based principles: internal models that allow an agent to compare policy checkpoints, monitor its learning capacity, and represent its current objective; and external models that learn environment dynamics and the behavior of other agents. We anticipate that discussion on these principles will facilitate the design of algorithms, architectures, and benchmarks that explicitly evaluate model-based components in CMARL.
Show more
View full details