探索全球前沿学术脉络

01.

arXiv (CS.AI) 2026-06-18 DOI: arXiv:2409.03500

Quality Perceptions and Intended Engagement in Response to AI-Generated and AI-Assisted News

作者:

Fabrizio Gilardi↗Sabrina Di Lorenzo↗Juri Ezzaini↗Beryl Santa↗Benjamin Streiff↗Eric Zurfluh↗Emma Hoes↗

arXiv:2409.03500v4 Announce Type: replace-cross Abstract: The increasing use of artificial intelligence (AI) in news production raises important questions about how audiences perceive and respond to AI-generated journalism. This preregistered survey experiment (N = 599, German-speaking Switzerland) examines (i) perceptions of article quality (measured as credibility, readability, and expertise) across news excerpts that were human-written, AI-assisted, or fully AI-generated, and (ii) self-reported intentions to engage following disclosure of AI involvement. Participants rated two short news excerpts before learning how they had been produced. Articles across all conditions were evaluated similarly in perceived quality. After disclosure, participants in the AI-assisted and AI-generated conditions reported a higher willingness to continue reading their assigned articles compared to the control group, but future willingness to read AI-generated news did not differ across conditions. Overall, the findings suggest that readers assess AI-generated and human-written news comparably in quality, while disclosure of AI use can momentarily increase curiosity or interest without yet changing longer-term reading intentions.

阅读与讨论 → 访问原文 →

02.

arXiv (CS.LG) 2026-06-15 DOI: arXiv:2606.14259

Beyond a Single Explanation of the Adam–SGD Gap

作者:

Chenxiang Zhang↗Rustem Islamov↗Enea Monzio Compagnoni↗Jun Pang↗Aurelien Lucchi↗Antonio Orvieto↗

arXiv:2606.14259v1 Announce Type: new Abstract: Prior work has identified several factors that can contribute to the performance gap between Adam and SGD, spanning data aspects, architecture design, and optimization properties. Yet these explanations are often studied in isolation, leaving their relative importance unclear. In this work, we revisit these hypotheses through a controlled empirical study across vision, language, genomics, and graph tasks, spanning modern and classical architectures, and carefully designed training setups. Our results suggest that no single factor consistently explains the Adam–SGD gap. For instance, the Adam advantage can (1) persist under a uniform vocabulary distribution yet nearly disappear under a heavy-tailed one; (2) reverse in favor of SGD in softmax-attention models; and (3) become larger under soft architectural modifications, e.g., when ReLU is replaced by a GeLU nonlinearity. This suggests that the gap arises from nontrivial data and architecture interactions, rather than from a single common factor. Yet, we observe a pattern across our settings: a crossover batch size at which the relative advantage shifts from SGD to Adam as the batch size scales. These empirical results are captured by our theoretical gap model, which predicts this batch-size-dependent crossover. Our perspective helps reconcile several existing hypotheses while offering practical insights across domains.

阅读与讨论 → 访问原文 →

03.

arXiv (CS.CV) 2026-06-17 DOI: arXiv:2606.17362

DriveJudge: Rethinking Autonomous Driving Evaluation with Vision-Language Models

作者:

Xinglong Sun↗Kevin Xie↗Jenny Schmalfuss↗Despoina Paschalidou↗Xiuming Zhang↗Sanja Fidler↗Kashyap Chitta↗Jose M. Alvarez↗

Autonomous driving has shifted towards end-to-end policy learning, where reliable, interpretable policy evaluation is a fundamental challenge as driving quality is highly context-dependent. Commonly used rule-based driving metrics like EPDMS are interpretable but lack context-awareness, while recent VLMbased evaluations are context-aware but limited by ambiguous VLM outputs and weak physical grounding. To evaluate driving in a manner that is both interpretable and context-aware, we introduce DriveJudge. DriveJudge is a driving evaluation agent that combines rule-grounded evaluation with Vision-Language Model (VLM) reasoning and selectively invokes physically-grounded deterministic rule functions after interpreting the environmental context. To train and evaluate DriveJudge, we curate a large-scale dataset of 33,577 challenging driving samples with human annotations on whether the driving behavior is reasonable in the given scenario. With this dataset, we address the underexplored problem of driving metric evaluation, and introduce two human-aligned benchmark tasks: Driving Quality Classification and Trajectory Preference Selection. DriveJudge outperforms EPDMS for driving quality classification by 21.23 AUC, and the recent VLM-based DriveCritic for trajectory preference selection by 6.5%, setting a new standard for interpretable and precise driving evaluation.

阅读与讨论 → 访问原文 →

04.

medRxiv (Medicine) 2026-06-22 DOI: HASH:27e222c43fb222f9f6728da3d9ef5ed1

A Controlled Human Malaria Infection model for relapsing Plasmodium vivax

作者:

Duncan↗A. D. S↗Geraedts↗T. J. M↗Manlutac↗Baker↗E. C↗van Heerden↗J. K↗Roberts↗T. W↗Rigby↗…

Background Plasmodium vivax malaria relapses are a major source of morbidity and onward transmission of infection. The underlying mechanisms are poorly understood and current therapies sub-optimal. We examined the safety and feasibility of a controlled human malaria infection (CHMI) model for relapsing P. vivax. Methods We conducted an open-label, proof-of-concept, CHMI study of relapsing P. vivax. Healthy, malaria-naive, Duffy-positive adults aged 18-45 years with extensive CYP2D6 metaboliser phenotype and normal blood glucose-6-phosphate dehydrogenase (G6PD) levels were recruited in Oxford, UK. Mosquito-bite CHMI was performed in Nijmegen, The Netherlands, using Anopheles stephensi mosquitoes infected with PvW1, a clonal isolate of P. vivax from Thailand. All follow-up visits were conducted in Oxford, UK. Primary P. vivax infections (qPCR > 500 genome copies/mL) were treated with artemether-lumefantrine (80mg/480mg at 8, 24, 36, 48 and 60 hours). From Day 28 following CHMI, participants attended a fortnightly clinic for clinical review and qPCR blood sampling, with additional assessments performed for any reported symptoms. P. vivax relapse infections (qPCR > 500 genome copies/mL) were treated with artemether-lumefantrine as per primary infection. Definitive anti-malarial treatment with atovaquone-proguanil (1000mg/400mg once daily for three days) and primaquine (0{middle dot}5 mg/kg/day for 14 days) was administered six months following CHMI, regardless of parasitaemia or symptoms. The primary objective was to assess the safety, feasibility and frequency of relapsing P. vivax after CHMI. Remote follow-up (5 years) is ongoing. The study is registered with ISRCTN registry (ISRCTN48625883). Findings 20 participants were screened for eligibility from 21 January 2025. Five participants (median age 22 years) underwent CHMI (five infected mosquitoes per participant) on 15 April 2025. All participants developed primary P. vivax infection and experienced at least one relapse infection. Two participants experienced a second relapse. Overall incidence rate was 3{middle dot}6 relapse infections per person-year. Solicited adverse events were mild or moderate and there were no serious adverse events. Definitive anti-malarial treatment was administered to all participants. One participant experienced primaquine-induced methaemoglobinaemia, resolving with early discontinuation of treatment (total dose 5{middle dot}3 mg/kg). To date, more than six months after primaquine treatment, no further relapses have been recorded. Interpretation CHMI of relapsing P. vivax is safe and feasible, allowing exploration of the mechanisms underlying relapse infections and providing a platform for future anti-relapse efficacy studies. Funding European Union Horizon Europe programme and UK Research and Innovation (UKRI) via OptiVivax consortium; UK National Institute for Health and Care Research Biomedical Research Centre: Oxford; and UK Medical Research Council.

阅读与讨论 → 访问原文 →

05.

arXiv (CS.LG) 2026-06-12 DOI: arXiv:2606.10678

One Step Closer to Ground Truth: A Multi-Scale Residual-Aware Representation Learning Pipeline for Predicting Time Series Data

作者:

Amrijit Biswas↗Mustafa Kamal↗Robin Krambroeckers↗M. M. Lutfe Elahi↗Sifat Momen↗Nabeel Mohammed↗Shafin Rahman↗

arXiv:2606.10678v2 Announce Type: replace Abstract: Transformer-based models have emerged as leading paradigms in time-series forecasting in recent years, employing self-attention mechanisms to capture long-range dependencies. Despite their success, these single-stage forecasting architectures exhibit persistent systematic residual biases arising from structural discrepancies, unmodeled stochastic components, or inadequate multi-scale temporal representations. This limitation persists when residuals are treated as irreducible noise, precluding adaptive correction of structured error patterns. To address this limitation, we introduce a two-stage, model-agnostic framework that explicitly decouples forecasting and residual learning into distinct stages of representation learning. A base transformer first generates the initial predictions. Subsequently, a dedicated meta-corrector dynamically models structured error patterns across multivariate channels, preserves cross-variable dependencies, and iteratively refines the residual bias of the base transformer. By formalizing this pipeline as a hypothesis space expansion, our framework addresses approximation limitations inherent in single-stage architectures, removes reliance on restrictive assumptions, and enables end-to-end learning of complex error dynamics. Evaluated on eight popular benchmark datasets using established protocols, our approach achieves state-of-the-art performance, with significant improvements in standard metrics (MSE, MAE). The results demonstrate the framework's ability to mitigate systematic biases and enhance robustness to complex temporal dynamics, advancing the practical applicability of transformer-based forecasting models.

阅读与讨论 → 访问原文 →

06.

arXiv (CS.LG) 2026-06-16 DOI: arXiv:2606.16693

Learning Hybrid Biophysical Neuron Models with Neural ODEs

作者:

Jonas Beck↗Michael Deistler↗D\'ora Vikt\'oria Moln\'ar↗Jakob H. Macke↗Philipp Berens↗

arXiv:2606.16693v1 Announce Type: cross Abstract: Biophysical neuron models link measurements of neural activity to underlying cellular mechanisms. Yet, a central challenge is that the kinetics of many ion channels are poorly characterized, and practical simplifications – omitting channels or reducing morphological detail – introduce systematic gaps between model and biology. Bridging these gaps requires approaches that can flexibly discover unmodeled dynamics while preserving mechanistic interpretability. Here, we introduce a hybrid modeling framework that embeds neural ordinary differential equations into conductance-based biophysical models to capture unknown currents or mis-specified channel kinetics. By parameterizing the neural ODE in terms of voltage-dependent steady-state and time-constant functions, we recover interpretable gating dynamics directly from voltage recordings without assuming a functional form. We show that the hybrid model fits the gating kinetics of 2400 ion channel models and recovers unknown gating dynamics from single current-clamp recordings, generalizing to out-of-distribution stimulus regimes under realistic inputs and parameter misspecification. We also use our method to reduce a multicompartment model of a cortical neuron into a single-compartment hybrid model with a learned axial current, yielding up to an order of magnitude lower computational cost. Together, our results establish a plug-and-play framework for selectively replacing unknown components of conductance-based models with neural ODEs while preserving their mechanistic structure.

阅读与讨论 → 访问原文 →

07.

arXiv (CS.LG) 2026-06-15 DOI: arXiv:2606.14245

Where Black-box Drug-Target Interaction Prediction Models Look: Cross-Method Explainability

作者:

Ali Vefghi↗Zahed Rahmati↗Mohammad Akbari↗

arXiv:2606.14245v1 Announce Type: new Abstract: Drug-target interaction (DTI) and affinity (DTA) predictors increasingly achieve strong benchmark scores, yet their internal use of sequence, fingerprint, and graph features often remains opaque. We present an interpretability audit of BridgeDPI architecture on three different datasets including Gao, Human, and C.elegans. This study combines gradient-based attributions – integrated gradients, saliency, layer-wise relevance propagation, SmoothGrad, and SmoothGrad-IG – with feature-wise occlusion ablation and strict intersection consensus across methods to reduce single-explainer bias. We summarize sensitivity and signed effects at raw inputs, at the bridge similarity scaffold, and through the graph convolution, including edge-level sensitivities and targeted edge removals. The results show that explainability is most informative when treated as model criticism: it reveals modality dominance, padding and special-token artifacts, dataset-dependent cooperative versus suppressive effects across layers, and chemistry-consistent fragment and composition motifs where methods agree. These analyses do not substitute for structural or experimental ground truth, yet they can provide testable hypotheses for downstream validation in computational drug discovery pipelines. More broadly, applying modern XAI to contemporary DTI/DTA models is still an early pass over the rich structure implicit in trained weights and data – yet even this first layer of scrutiny already helps researchers relate predictions to drug- and target-side representations and to prioritize external validation.

阅读与讨论 → 访问原文 →

08.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2512.19011

Do You Really Need a GPU to Guard Your LLM? CPU-Class Classifiers and Multi-Stage Pipelines for Safety Enforcement at Scale

作者:

Vasudev Majhi↗Dhruv Gupta↗Advait Singh↗Matthew Barker↗Dhruv Kumar↗

Safety classifiers that screen LLM inputs for jailbreak attempts have become standard deployment components, yet almost all production systems rely on GPU-based models: fine-tuned transformers and LLM-as-a-judge pipelines. These approaches impose significant per-query latency and infrastructure cost. Very little research has asked whether CPU-based classifiers, such as support vector machines and gradient-boosted trees trained on TF-IDF features, can match their accuracy across the conditions that production deployments encounter. We evaluate five CPU classifier families, Mamba-130M as an SSM-based GPU classifier, and transformer-based GPU models (DeBERTa-v3 and Gemma-2B with LoRA) across nine jailbreak sources and three regimes: in-distribution (D1), out-of-distribution (D2), and adversarially obfuscated (D3). On D1, the best CPU classifier matches the best transformer GPU model at roughly one-fifth the deployment cost. On D2, CPU classifiers fail via confident miscalibration, producing high-confidence false negatives that bypass escalation entirely. On D3, CPU classifiers outperform transformer GPU models by more than 26 percentage points in F1. Based on these complementary failure modes, we design GuardChain, a three-stage safety pipeline (Regex -> CPU -> GPU) that routes each prompt to the cheapest stage capable of a confident decision. The CPU stage alone resolves 80\% of in-distribution prompts at near-peak accuracy, and the GPU stage recovers the out-of-distribution failures. For practitioners deploying LLM safety at scale, this work provides evidence that GPU-class infrastructure is unnecessary for the majority of traffic.

阅读与讨论 → 访问原文 →

09.

arXiv (CS.CV) 2026-06-11 DOI: arXiv:2606.11289

i1: A Simple and Fully Open Recipe for Strong Text-to-Image Models

作者:

Boya Zeng↗Tianze Luo↗Shu Pu↗Jucheng Shen↗Taiming Lu↗Gabriel Sarch↗Zhuang Liu↗

Diffusion models have consistently driven progress in text-to-image generation. However, it is challenging to attribute recent progress to specific modeling and data choices: state-of-the-art open-weight models provide limited ablations, and do not disclose their training data and full training details. The research community needs fully open (weights, data, and code) models as a foundation for further research; yet existing fully open models still fall significantly short of leading models in performance. In this project, we conduct a systematic investigation of the modeling and data design choices in text-to-image diffusion training and inference with 300+ controlled experiments totaling 700K+ TPU v6e hours. Our experiments highlight several empirical findings (e.g., equal weighting is a strong default for mixing curated datasets) and simple design decisions (e.g., larger text encoder adapters improve performance with minimal added parameters) for training strong models. Guided by these insights, we train i1, a 3B-parameter text-to-image diffusion model using only publicly available datasets. i1 is competitive with leading models on five representative benchmarks (GenEval, DPG, PRISM, CVTG-2K, and LongText), and outperforms the best existing fully open model by 29.5 absolute percentage points on average. We provide the i1 checkpoints, training and inference code, and the data processing pipeline. Together, our findings and the i1 recipe establish a practical foundation for future open research in text-to-image diffusion models. Our code is available at https://github.com/zlab-princeton/i1.

阅读与讨论 → 访问原文 →

10.

arXiv (CS.AI) 2026-06-11 DOI: arXiv:2606.11794

Multimodal Ordinal Modeling of Alzheimer's Disease Severity Using Structural MRI and Clinical Data

作者:

Boris-Stephan Rauchmann↗Jonathan Laib↗Buse Ercik↗Robert Perneczky↗Sergio Altares-L\'opez↗

arXiv:2606.11794v1 Announce Type: cross Abstract: Neurodegenerative diseases such as Alzheimer's disease (AD) require accurate and scalable tools for assessing disease severity, yet current clinical staging remains time-intensive and prone to variability. We propose an attention-enhanced multimodal machine learning framework with ordinal regression for automated and interpretable AD severity staging. The framework integrates T1-weighted MRI with demographic and genetic variables and compares unimodal and multimodal architectures using ordinal and non-ordinal prediction heads. Models were trained and validated using cohort-stratified splits derived from the ADNI, AIBL, and NIFD datasets. A strictly held-out test set was constructed using subjects excluded from all training, validation, preprocessing, and hyperparameter tuning procedures, with subject-level splitting employed throughout to prevent data leakage. Among unimodal approaches, the T1-weighted MRI model achieved slightly higher adjacent-stage accuracy (0.963) and agreement with clinical staging (QWK 0.444) than the tabular model (QWK 0.433). Integrating imaging, demographic, and genetic information improved overall performance. The multimodal non-ordinal baseline achieved the lowest prediction error (MAE 0.340), whereas the ordinal multimodal model achieved the highest adjacent-stage accuracy (0.970) and strongest agreement with clinical staging (QWK 0.549). These findings indicate that ordinal formulations better capture the ordered structure of the CDR scale and yield predictions more consistent with clinical staging. Explainability analyses using Grad CAM++ and SHAP demonstrated anatomically and clinically plausible model behavior, supporting transparent decision-making. Overall, attention-based multimodal learning with ordinal regression represents a robust, interpretable, and scalable approach for automated AD severity staging and AI-assisted clinical decision support.

阅读与讨论 → 访问原文 →

11.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2509.21862

Shachi: A Modular, Controllable Framework for LLM-Based Agent-Based Modeling of Emergent Collective Behavior

作者:

So Kuroki↗Yingtao Tian↗Kou Misaki↗Takashi Ikegami↗Takuya Akiba↗Yujin Tang↗

arXiv:2509.21862v3 Announce Type: replace Abstract: How collective behaviors emerge from the interactions of individual LLM-driven agents is a central question in artificial life, yet controlled study of these emergent dynamics has been hindered by the lack of a principled simulation framework for systematic experimentation. To address this, we introduce Shachi, a principled methodology and modular framework that decomposes an agent's cognition into core components: Configuration for intrinsic identity, Memory for contextual continuity, and Tools for extended capabilities, all orchestrated by an LLM reasoning engine. This decomposition treats each cognitive component as an independently controllable variable, enabling perturbation studies that trace how micro-level cognitive traits propagate into population-level dynamics. We investigate behavioral patterns across a 10-task benchmark spanning three levels of collective complexity. Shachi enables memory transfer across environment transitions, producing history-dependent behavioral shifts, and allows agents to simultaneously inhabit multiple environments, revealing cross-environment interference invisible in single-environment studies. Furthermore, in a real-world U.S. tariff shock case study, locally interacting agents with individually controlled cognitive components produce macro-level market dynamics directionally consistent with observed real-world outcomes. Our work provides a rigorous, open-source simulation framework for LLM-based ABM, aimed at fostering cumulative scientific inquiry into the emergent collective behaviors of interacting artificial agents.

阅读与讨论 → 访问原文 →

12.

arXiv (CS.LG) 2026-06-15 DOI: arXiv:2606.14235

Implicit Variational Rejection Sampling

作者:

Jian Xu↗Shigui Li↗Wei Chen↗Jiacheng Li↗Zhiqi Lin↗Delu Zeng↗Xinghao Ding↗John Paisley↗Qibin Zhao↗

arXiv:2606.14235v1 Announce Type: new Abstract: Variational Inference (VI) is a fundamental inference technique in Bayesian machine learning for approximating complex posterior distributions. Traditional VI often relies on the mean-field factorization, which can inadequately capture true posterior complexity. Recent advancements have leveraged neural networks to model implicit distributions, offering increased flexibility. However, the practical constraints of neural network architectures still produces inaccuracies. In this paper, we propose a method called Implicit Variational Rejection Sampling (IVRS), which integrates implicit distributions with rejection sampling to improve the posterior approximation. Our method uses neural networks to construct implicit proposal distributions, and rejection sampling with a discriminator network that estimates the density ratio between the implicit proposal and the true posterior for refining the approximation. Towards this end, we introduce the Implicit Resampling Evidence Lower Bound (IR-ELBO) as a metric to characterize the resampled distribution's quality and derive a tighter variational lower bound. Experimental results demonstrate that our method outperforms traditional variational inference techniques.

阅读与讨论 → 访问原文 →

13.

arXiv (CS.LG) 2026-06-19 DOI: arXiv:2606.19603

Comparing Linear Probes with Mahalanobis Cosine Similarity

作者:

Zhuofan Josh Ying↗Peter Hase↗Nikolaus Kriegeskorte↗

arXiv:2606.19603v1 Announce Type: new Abstract: Linear probes are widely used in interpretability research and often compared by cosine similarity. The Mahalanobis cosine similarity (MCS) between two directions, which reweights the inner product by test data covariance, is a natural task-aware refinement. Ying et al. (2026) report that a probe's MCS to a reference probe trained on the out-of-distribution (OOD) data near-perfectly linearly predicts the probe's OOD AUROC (R^2 = 0.98). Here, we extend this empirical finding across models, layers, and concept domains, and prove this general phenomenon in closed form: For balanced classes whose projections are Gaussian, OOD AUROC and MCS to the reference probe are linear because both are sigmoid-shaped functions of the probe's signal-to-noise ratio (SNR) on the test data. The theory also predicts when this linearity fails, which we verify empirically. MCS offers a theoretically grounded and empirically effective alternative to Euclidean cosine similarity for comparing linear probes.

阅读与讨论 → 访问原文 →

14.

arXiv (CS.AI) 2026-06-18 DOI: arXiv:2604.25848

A Distributionally Robust Reinforcement Learning Framework for Constrained Urban EV Dispatch

作者:

An Nguyen↗Hoang Nguyen↗Phuong Le↗Hung Pham↗Cuong Do↗Laurent El Ghaoui↗

arXiv:2604.25848v2 Announce Type: replace Abstract: We study city-scale control of electric-vehicle (EV) ride-hailing fleets where dispatch, repositioning, and charging decisions must respect charger and feeder limits under uncertain, spatially correlated demand and travel times. We formulate the problem as a hex-grid semi-Markov decision process (semi-MDP) with mixed actions – discrete actions for serving, repositioning, and charging, together with continuous charging power – and variable action durations. To guarantee physical feasibility during both training and deployment, the policy learns over high-level intentions produced by a masked, temperature-annealed actor. These intentions are projected at every decision step through a time-limited rolling mixed-integer linear program (MILP) that strictly enforces state-of-charge, port, and feeder constraints. To mitigate distributional shifts, we optimize a Soft Actor-Critic (SAC) agent against a Wasserstein-1 ambiguity set with a graph-aligned Mahalanobis ground metric that captures spatial correlations. The robust backup uses the Kantorovich-Rubinstein dual, a projected subgradient inner loop, and a primal-dual risk-budget update. Our architecture combines a two-layer Graph Convolutional Network (GCN) encoder, twin critics, and a value network that drives the adversary. Experiments on a large-scale EV fleet simulator built from NYC taxi data show that PD-RSAC achieves the highest net profit, reaching \$1.22M, compared with \$0.58M-\$0.70M for strong heuristic, single-agent RL, and multi-agent RL baselines, including Greedy, SAC, MAPPO, and MADDPG, while maintaining zero feeder-limit violations.

阅读与讨论 → 访问原文 →

15.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.15057

AutoDojo: Adaptive Attacks Expose Superficial Defenses and User-Underspecification Limits in LLM Agents

作者:

Xinhang Ma↗Taoran Li↗Chaowei Xiao↗Zhiyuan Yu↗Ning Zhang↗Yevgeniy Vorobeychik↗

arXiv:2606.15057v1 Announce Type: cross Abstract: Indirect prompt injection (IPI) is a major security threat to LLM-powered agents. Thus, a growing body of work have proposed a variety of defensive approaches against IPI. These can be grouped into three broad categories: 1) prompt-based (using prompting as a way to prevent agents from following malicious instructions), 2) detection-based (identifying and filtering malicious instructions), and 3) system-level (using systems insights, such as control and data isolation, for defense). However, commonly used benchmarks for evaluating defense, such as AgentDojo, are inherently static, generating a fixed distribution of IPI attacks. Consequently, static benchmarks do not usefully evaluate defense robustness to adaptive threats. We address this issue by developing AutoDojo, an adaptive extension of AgentDojo that optimizes IPI against a given defense. Using AutoDojo against state-of-the-art IPI defenses across three task suites and five target models, we make two key observations. First, many defenses offer only limited protection: a cheap, black-box adaptive attack using a frontier LLM to iteratively optimize the injection raises attack success rate (ASR) well above the level achieved by static injections against nearly all evaluated defenses. Against a filter that reduces static ASR to 0\%, AutoDojo recovers 28\% overall and 64\% on action-open tasks. Second, for prompt-level and filter-based defenses, ASR is substantially higher on action-open tasks – where the user's request delegates the action itself to attacker-controlled content – than on precisely specified tasks. This is a structural limit: on such tasks the injection can pose as ordinary data rather than an explicit instruction, bypassing defenses that rely on detecting instruction-like text. AutoDojo is publicly available at https://github.com/xhOwenMa/AutoDojo.

阅读与讨论 → 访问原文 →

16.

arXiv (CS.CL) 2026-06-19 DOI: arXiv:2606.04075

Large Language Models Hack Rewards, and Society

作者:

Wei Liu↗Xinyi Mou↗Hanqi Yan↗Zhongyu Wei↗Yulan He↗

Reinforcement learning (RL) has become a dominant post-training paradigm, enabling large language models (LLMs) to learn from rewards. We observe that societal regulations are structurally similar to reward functions. They define measurable outcomes, thresholds, and exceptions, while often leaving institutional intent only partially specified. We hypothesise that the RL training process may exploit these gaps and therefore ask whether models' well-known tendency to hack reward functions during RL can scale into a more consequential failure mode named societal hacking: discovering loopholes in the rules society runs on. To study this phenomenon, we introduce SocioHack, a sandbox of 72 societal environments, and find that within these environments, reward hacking naturally emerges and leads to regulatory loophole discovery. Models learn to hack the social rules and generate strategies that remain technically compliant while defeating regulatory intent, and current LLM safeguards provide only limited mitigation. Therefore, collecting in-the-wild feedback for model training requires greater caution, and we need a next-generation post-training paradigm for safely iterating LLMs in real society.=

阅读与讨论 → 访问原文 →

17.

Nature (Science) 2026-06-10 DOI: HASH:aac7517d04bfd73b8fd4bf97a3207e83

Mitochondria tethered to the nucleus secure its energy supply

作者:

Philippa Clarke↗

Direct interactions between the cell’s powerhouses and nuclear pores might channel energy straight into the nucleus, fuelling cell division and differentiation. Direct interactions between the cell’s powerhouses and nuclear pores might channel energy straight into the nucleus, fuelling cell division and differentiation.

阅读与讨论 → 访问原文 →

18.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.16541

The Faithfulness Gap: Certifying Semantic Equivalence Between Natural-Language and Formal Mathematical Statements

作者:

Noor Islam S. Mohammad↗Tamim Sheikh↗

arXiv:2606.16541v1 Announce Type: new Abstract: Autoformalization, translating natural-language mathematics into formal proof assistants, is bottlenecked not by translation fluency but by faithfulness: a formal statement can typecheck and be provable, yet still encode a different theorem than the source intended. We introduce Bidirectional Provability Fingerprinting (\bpf{}), a framework that certifies faithfulness by characterizing each candidate through its forward and backward consequence neighborhoods in the ambient theory and matching these against probes derived from the natural-language statement. We further introduce four novel components: (i) Counterfactual Probe Generation (\cpg{}), a contrastive procedure that synthesizes probes targeting specific drift directions; (ii) the Equivalence Spectrum, a continuous faithfulness score that replaces brittle binary verdicts; (iii) Adaptive Probe Budget Allocation (\apba{}), an information-theoretic budget router; and (iv) Faithfulness-Guided Decoding (\fgd{}), which uses \bpf{} signals as a reward during autoformalization. We prove a drift detection theorem and a PAC-faithfulness result establishing that the equivalence class of a natural language statement is learnable from $\mathcal{O}(\log(1/\delta)/\varepsilon)$ probes under mild assumptions. We release \driftbench{}, a benchmark of $2{,}183$ NL/Lean~4 pairs with controlled drift labels across six subfields of mathlib4. \bpf{}\,+\,\cpg{} detects $89.6\%$ of drifted formalizations at a $3.0\%$ false-positive rate-against $41.2\%$ for typecheck and $63.3\%$ for LLM-judge baselines, and \fgd{} reduces the rate at which a state-of-the-art autoformalizer emits drifted statements by $47\%$. https://pmlrbd.github.io/BPF/

阅读与讨论 → 访问原文 →

19.

arXiv (quant-ph) 2026-06-12 DOI: arXiv:2606.12527

Statistical Mechanics and Symmetries of Non-Abelian Anyon Proliferation: From Deformation to Decoherence

作者:

Avi Vadali↗Robijn Vanhove↗Ruben Verresen↗Jason Alicea↗Pablo Sala↗

arXiv:2606.12527v1 Announce Type: new Abstract: Topological quantum computation relies on braiding non-Abelian anyons, but requires the underlying topological order to survive imperfect state preparation and environmental noise. We show that the instability of topological order to wavefunction deformations and to decoherence, with the latter probed by syndrome distributions, are generically captured by stat-mech models whose symmetries naturally expose the corrupting anyonic excitations. As an example, we combine this framework with Monte-Carlo simulations to resolve the stability of $D_4$ topological order under deformations and quantum channels that proliferate multiple non-Abelian anyon species that individually are unable to condense. We show that beyond a finite threshold, proliferation of two non-Abelian anyon species parasitically condenses a shared Abelian-anyon fusion outcome, destroying the topological order. Our symmetry-based approach sharply differentiates the resulting trivial phase from that obtained by condensing all Abelian charges; in other words, the trivial phase "remembers" which anyons condensed. This framework provides a first step into identifying the relevant symmetry for optimal decoders, conditioned on syndrome measurements, of non-Abelian topological order.

阅读与讨论 → 访问原文 →

20.

arXiv (CS.LG) 2026-06-12 DOI: arXiv:2606.12966

Circuit Synchronization Precedes Generalization: Causal Evidence from Fourier Structure in Grokking Transformers

作者:

Achyuthan Sivasankar↗

arXiv:2606.12966v1 Announce Type: new Abstract: Grokking – where a transformer on modular arithmetic suddenly transitions from near-chance to near-perfect validation accuracy – is attributed to a Fourier circuit, but its timing, causal structure, and controllability remain poorly understood. We introduce the Frequency Synchronization Degree (FSD), a normalised, permutation-tested metric for Fourier circuit synchronisation requiring no prior circuit knowledge. Across nine modular addition configurations (primes p in {53, 71, 97, 113, 131}, three seeds), FSD synchronises 500-3,000 steps before grokking (mean lead +1,722 steps; all nine positive, sign-test p~0.004), and precedes a restricted-logit loss baseline (Nanda et al.'s excluded loss) in all nine cases, making it the earliest available predictor. We provide direct causal evidence that the inter-phase gap is a regularisation phenomenon: forking training at the FSD-ceiling step and varying weight decay lambda produces strictly monotone earlier grokking, with Delta_t proportional to 1/lambda. This law replicates across three primes (p in {53,97,131}; R^2=1.00 and R^2=0.99 for two clean cases), captured as Delta_t ~ C/lambda, consistent with (1/lambda)*log(||W_mem||/tau). Architecture ablations show an attention-only model groks with a strong FSD precursor; an MLP-only model never groks; a single-layer model's FSD lags, confirming the precursor is a multi-block circuit property.

阅读与讨论 → 访问原文 →

21.

arXiv (CS.CL) 2026-06-18 DOI: arXiv:2606.18856

Approximate Structured Diffusion for Sequence Labelling

作者:

Nicolas Floquet↗Joseph Le Roux↗Nadi Tomeh↗

Sequence labelling, a core task of Natural Language Processing (NLP), consists in assigning each token of an input sentence a label. From a Machine Learning point of view, sequence labelling is often cast as a Linear-Chain Conditional Random Field (CRF) parametrised by a neural network. While this approach gives good empirical results, CRFs assume a finite decision span (eg label bigrams) which can limit their expressivity and hurt performance when long-range dependencies are required. We show we can leverage diffusion to train a CRF conditioned on an entire label sequence, with the caveat that the condition is on a noisy version of labels. We show experimentally that this method, in conjunction with approximate CRF inference, improves label accuracy with a 16.5% error reduction for POS-tagging.

阅读与讨论 → 访问原文 →

22.

bioRxiv (Bioinfo) 2026-06-11 DOI: HASH:a1ad26c4fc4beb0b2feabce024e49493

Viability of engineered AAVs via protein language models

作者:

Desrosiers↗Ocari↗Trinquier↗Zin↗E. A↗Van Meter↗Delmas↗Tekinsoy↗Planul↗Nemoto↗Dalkara↗Ferrari↗…

Capsid engineering has greatly improved the performance of recombinant AAV vectors used for gene therapy. One commonly used strategy is the insertion of a short, 7-mer, peptide into surface-exposed loops to modify receptor interactions and enhance cell entry. While effective in receptor retargeting and improved transduction, these insertions might destabilize the capsid protein, hinder assembly, and thus limit production. While previous attempts have used deep mutational scanning and AI to predict which insertions are viable, there is lack in understanding the structural consequences of these peptide insertions at the amino-acid level. Here we combined experiments, deep sequencing and large protein language models to gain insight on the impact of 7-mer insertions on the VR-VIII region. We first characterize the biochemical properties of viable insertions, thus identifying which residues are well tolerated, and which should instead be avoided. We then focus on the nearby context of those insertions, by studying the effect of the linkers, either for highly diverse libraries or for individual variants known for their efficiency. Next, we study the broader context, by extending our analysis to the whole capsid sequence, and identifying regions that can tolerate insertions without long-ranged structural deformations that could affect capsid functionality. We conclude with a cross-serotype comparison and a viability analysis of tens of previously engineered variants. Our work showcases how AI can uncover structure-function rules governing the success of engineered AAV capsids.

阅读与讨论 → 访问原文 →

23.

arXiv (quant-ph) 2026-06-17 DOI: arXiv:2606.17264

Coupled-Mode Equations with Arbitrary Mode Combinations for Kinetic-Inductance Superconducting Traveling-Wave Parametric Devices: Theory and Experimental Validation

作者:

F. Patricio Mena↗Camilo Espinoza↗Ryan O. Berriel↗Ricardo Finger↗David J. Thoen↗

arXiv:2606.17264v1 Announce Type: cross Abstract: The coupled-mode equations (CMEs) have proven very successful in describing parametric processes in nonlinear optics. More recently, the same formulation has been used to model microwave superconducting parametric amplifiers and frequency multipliers. However, when applied to the microwave regime, not all assumptions remain valid and losses play a more dramatic role. Here, we revisit the CMEs applied to traveling-wave superconducting amplifiers to include losses and provide a formulation that enables their systematic derivation for any combination of traveling waves. As examples, we discuss the impact of unwanted harmonics and intermodulation products on parametric amplification, as well as harmonic generation. We verify that, if not properly accounted for, device performance can deviate considerably from the ideal case. Furthermore, using a superconducting CPW-based artificial transmission line and combining an independent experimental determination of its nonlinear parameter $I'_*$ with simulations of its linear properties, we obtain a parameter-free validation of this formulation. The nonlinear parameter was determined to be $I'_* \approx 27$ mA which, surprisingly, scales with the theoretical depairing current and not with the much smaller critical current of the device. For the validation, we measured multiple-harmonic generation and found excellent agreement between theory and experiment. The fact that $I'_* \gg I_C$ has direct implications for device design.

阅读与讨论 → 访问原文 →

24.

arXiv (CS.CL) 2026-06-17 DOI: arXiv:2606.17890

Dynamic Rollout Editing for Reducing Overthinking in RL-Trained Reasoning Models

作者:

Zihao Wei↗Wenjie Shi↗Liang Pang↗Jingcheng Deng↗Shicheng Xu↗Shasha Guo↗Zenghao Duan↗Jiahao Liu↗Jingang Wang↗Huawei Shen↗Xueqi Cheng↗

Long-form chain-of-thought reasoning can improve LLM performance on complex tasks, but models often continue generating unnecessary reasoning after a correct answer has emerged. We refer to this behavior as overthinking. We study this phenomenon from the perspective of GRPO-style reinforcement learning (RL) post-training, framing it as a training-time credit-assignment problem rather than merely a decoding-time stopping problem. In rollouts sampled at the onset of GRPO training, we observe that successful trajectories can exhibit a slightly higher degree of overthinking than unsuccessful trajectories for the same prompts. This early imbalance provides a starting point for an undesirable feedback loop: because GRPO assigns sequence-level credit, it cannot distinguish the solution-reaching prefix from the unnecessary continuation that lengthens a successful trajectory. Both receive positive update signal, allowing the initial imbalance to grow into more severe overthinking during training. To address this issue, we introduce Dynamic Rollout Editing (DRE), a training-time intervention for successful trajectories that continue thinking after answer emergence. DRE preserves the accepted verified prefix, edits the remaining thinking, and prefers the edited trajectory within the same RL group, weakening the preference signal for unnecessary thinking without penalizing the reasoning needed to reach the answer. Experiments across diverse tasks show the effectiveness of DRE.

阅读与讨论 → 访问原文 →

25.

arXiv (CS.CV) 2026-06-12 DOI: arXiv:2606.13497

SPARC: Reliable Spatial Annotations from Robot Demonstrations at Scale

作者:

Nils Blank↗Paul Mattes↗Maximilian Xiling Li↗Jakub Suliga↗Thomas Roth↗Moritz Reuss↗Pankhuri Vanjani↗Rudolf Lioutikov↗

This work introduces Spatial Annotations from Robot Demonstrations with Reliability Calibration (SPARC), a risk-aware framework that automatically labels robot demonstrations with structured spatial annotations and assigns each annotation a reliability score. Structured spatial annotations, such as bounding boxes, object trajectories, and manipulation phase labels, benefit a broad range of robotics applications from training grounded robot policies and embodied foundation models to motion planning and hierarchical task composition. Existing automated pipelines generate such annotations at scale but provide no reliable quality signal: detector confidence is poorly calibrated for annotation correctness, forcing a choice between accepting noisy labels or discarding useful samples. In contrast to existing automated pipelines, SPARC leverages the spatio-temporal structure inherent to robot tasks to generate a reliability signal, reducing noisy labels and retaining more useful samples. We further introduce Interaction-Aware Bench (IA-Bench), a benchmark that measures model accuracy in grounding the locations of interacted objects in robot demonstrations. On 1.7k human-annotated demonstrations spanning diverse embodiments and scenarios, SPARC significantly outperforms detection-only baselines in localization accuracy while retaining three times more samples at high-precision operating points. Our experiments demonstrate that models finetuned on our annotations achieve state-of-the-art results on object-grounding and pointing benchmarks among similarly sized models, while remaining competitive on broader spatial-reasoning suites without manually verified or annotated training data. Furthermore, policies trained on SPARC-generated annotations outperform baselines in cluttered, visually ambiguous real-world scenes. Code, data, and models are available at intuitive-robots.github.io/sparc-labeling.

阅读与讨论 → 访问原文 →