The paper introduces the first systematic robustness benchmark for local invisible image watermarking, evaluating five methods across 55 image transformations that include signal distortions, coordinate misalignments, indirect local edits, and direct watermark edits. Results show that all methods are vulnerable to some transformation, with MaskWM achieving the best payload recovery and localization but at the cost of image quality. The study highlights that robustness varies strongly with transformation type, especially noting that geometric misalignment and generative local edits can completely disrupt payload recovery.
By Kai Yao, Bence Szil\'agyi, Sebesty\'en Kamp, M\'at\'e Po\'or, M\'at\'e Szilveszter, Matyas K. Zsoldos, Marc Juarez
ECHO is a new paired diagnostic benchmark designed for Chinese full‑duplex spoken dialogue systems to evaluate context‑sensitive turn‑taking. It pairs examples that share the same overlap transcript but differ in preceding multi‑turn context, requiring either a Yield or Keep decision, and also includes off‑talk cases to test unnecessary yielding. The benchmark introduces pair accuracy, penalizing constant‑action policies, and shows that many systems are biased toward Yield, performing better on interruptions than on backchannels.
By Shuofeng Zhao, Hongwei Cai, Wenke Fan, Qingxiang Guo, Dawei Yang, Zhou Wang, Zhiyang Zhou, Yingxin Shang, Weixu Wang, Lin Yang, Shuran Zhou, Yang Song
arXiv:2609. 17435v1 Announce Type: new Abstract: We submit M\'eTRON-FR, a 125M GPT-2 pretrained on 92.
By Adam Zachary Wasserman, David Beauchemin
The paper presents a formal framework for majority‑minority learning tasks, demonstrating that standard gradient‑descent training can amplify bias by favoring majority groups and producing stereotypical predictors that ignore minority‑specific features. It identifies three key theoretical findings: the close proximity between full‑data and stereotypical predictors, the dominance of a region where training the entire model mainly captures majority traits, and a lower bound on the extra training needed to mitigate this effect. Experiments on deep learning for tabular and image classification illustrate these results.
By Fran\c{c}ois Bachoc (LPP), J\'er\^ome Bolte (TSE-R), Ryan Boustany (TSE-R), Jean-Michel Loubes (IMT, REGALIA)
The paper proposes Intelligent Interaction Techniques (IIxTs), arguing that current interaction techniques (IxTs) used in graphical user interfaces are largely static and were only modestly updated for smartphones. It suggests developing more intelligent IxTs that allow users to mix modalities—such as text, voice, and gestures—within a single interaction, thereby extending beyond the typical chat‑window approach seen in many AI interfaces. The proposal calls for research into new IxTs and the supporting infrastructure needed to realize these intelligent, multimodal interactions.
By Brad A. Myers
The paper examines whether removing declared language fields from de‑identified résumés eliminates demographic leakage in large language models. By keeping language attributes identical and varying only unstructured prose across five ethnocultural groups and three cue‑salience levels, the authors find that non‑language text still allows target‑group recovery (average 0.757, reaching 1.000 under high salience). They also show that evaluation design—such as allowing or forbidding ties—dramatically affects LLM‑as‑a‑judge outcomes, underscoring the importance of evaluation protocol in bias audits.
By Qiangju Chen, Yang Xiao
The paper introduces Adaptive Derivative-Ordered Random Explanation (ADORE), a unified framework that uses first- and second-order derivatives to capture nonlinear feature interactions and feature-sample dynamics. ADORE combines global feature importance with local sample contributions, quantifying both magnitude and direction of feature impact while identifying critical samples. It achieves computational efficiency via randomized SVD and dynamic sparsity detection, outperforming LIME and SHAP across tabular, text, and image data, and is released as an open-source Python package on GitHub.
By Lemen Chao, Ming Lei, Anran Fanga
arXiv:2609. 17464v1 Announce Type: cross Abstract: Multi-agent systems split a task across a tree of agents and justify the split with folklore: smaller contexts, cleaner separation, parallelism.
By Rong He
arXiv:2609.16603v1 Announce Type: new
Abstract: Full-context neural visual geometry is impractical for thousands of images, while sequence-based chunking poorly captures irregular non-local overlap i...
By Jeng Wen Joshua Lean, Ting-Yu Yen, Wei-Fang Sun, Simon See, Hung-Kuo Chu, Shih-Hsuan Hung
BenSparX introduces the first Bengali conversational speech dataset for Parkinson’s disease detection and pairs it with a robust, explainable machine learning framework. The framework uses diverse acoustic features, systematic feature selection, and advanced classifiers, achieving 95.67% accuracy, 95.62% F1, and 0.990 AUC. SHAP analysis is employed to explain feature contributions, and the model outperforms state‑of‑the‑art methods on other language datasets.
By Riad Hossain, Muhammad Ashad Kabir, Arat Ibne Golam Mowla, Animesh Chandra Roy, Ranjit Kumar Ghosh
BiCLIP is a bidirectional multimodal framework that enhances medical image segmentation by allowing visual features to iteratively refine textual representations, improving semantic alignment. It incorporates an augmentation consistency objective to stabilize learning against perturbed inputs. Experiments on QaTa-COV19 and MosMedData+ show that BiCLIP outperforms state‑of‑the‑art image‑only and multimodal baselines, achieving strong performance even with only 1% labeled data and resisting common clinical artifacts such as motion blur and low‑dose CT noise.
By Saivan Talaei, Fatemeh Daneshfar, Abdulhady Abas Abdullah, Mourad Oussalah
The paper reviews ethical and privacy risks of large language model (LLM)–enabled geospatial artificial intelligence (GeoAI), identifying eight recurring issues such as data provenance, spatial privacy, algorithmic bias, and technical risks. It evaluates current responses, noting many remain largely unaddressed or conceptual, and proposes a governance‑aware architecture with enforceable controls illustrated by a flood‑response routing scenario. The authors call for empirical validation, spatially specific interpretability tools, and workforce training to address these emerging risks.
By Maya Subramanian, Devika Jain
The paper titled "Position: AI Is Not Ready for Strategic Conflicts" argues that language‑model (LM) based open‑ended strategic wargames, while useful for simulating adversaries, institutions, and crisis response, pose significant safety risks. It identifies five failure modes—decision laundering, adjudication opacity, role collapse, escalation‑through‑adjudication, and failure of strategic imagination—and contends that such wargames should not inform real‑world planning or policy without an auditable safety case. Instead, the authors suggest using these simulations primarily as stress tests to expose potential failures in decision‑influencing LM agents.
By Mark Riedl, Glenn Matlin
The paper introduces a coverage-aware virtual IMU augmentation framework for human activity recognition. It selects diverse and scarce data points in a learned sensor embedding space, generates virtual IMU samples as prompts, ranks them by proximity and label consistency, and incorporates them into training with reliability-based weights. Experiments on public benchmarks demonstrate consistent performance gains over existing baselines, with ablation studies confirming the framework’s effectiveness.
By Jiayuan Gao, Yingwei Zhang, Ziyao Tang, Yuejia Ma, Yuanzhe Chen, Shuchao Song, Boshi Tang
The paper investigates whether natural language inference (NLI) can be performed using only interpretable, graph-based representations of evidence. It introduces a pipeline that decomposes sentences into atomic propositions, maps them to ConceptNet triples, and feeds three graphs—premise, hypothesis, and a retrieved ConceptNet subgraph—into a fine‑tuned 0.8‑billion‑parameter language model. On the SNLI dataset the graph‑only model reaches 89.7% accuracy, close to a text‑based baseline, while on ANLI it matches RoBERTa‑large on rounds R2 and R3 but lags on R1, illustrating a trade‑off between interpretability and performance.
By Younes Boufouss (LISN), Luc Pommeret (LISN, CNRS), Thomas Gerald (LISN), Patrick Paroubek (LISN, CNRS), Sophie Rosset (LISN, CNRS)
The paper proposes universal, tool‑based defenses for large language model agents that use external tools, addressing four types of adversarial attacks: direct and indirect prompt injection, memory poisoning, and backdoor attacks. Two main defenses are introduced: Attacker Tool Filtering, which uses anomaly detection to remove suspicious tools, and Normal Tool Recalling, which restores the agent’s original toolset before planning. The authors also add prompt‑based defenses such as Chain‑of‑Thought prompting and self‑reflection, and demonstrate that these methods dramatically lower attack success rates—often to 0%—across multiple open‑source and proprietary LLMs while maintaining or improving task performance.
By Xiaoyan Li, Yunli Wang
OPD‑Aha is a privileged on‑policy distillation method that improves multimodal reasoning by reconstructing the distillation target from the teacher’s isolated visual preference instead of relying on fragile teacher‑student discrepancies. It suppresses continuations that contradict the image, encouraging students to interrupt flawed reasoning with reflection tokens such as "wait" and "actually." This approach leads to consistent improvements across fine‑grained perception and complex multimodal reasoning benchmarks.
By Chenhao Qiu, Dawei Li, Yechao Zhang, Lei Gong, Zhen Tan
The paper evaluates five context‑trimming strategies for agentic large language model workflows, comparing them on metrics such as task success, protocol adherence, token savings, and latency. Conventional trimming methods save about 60% of tokens but achieve lower success rates, while protocol‑aware trimming raises success to 92.2% and adaptive guardrails further improve it to 96% success with 56% token savings. The study shows that preserving protocol‑critical state is more important than aggressive token removal, and that adaptive guardrails enhance efficiency, scalability, and reliability for long‑horizon agentic systems.
By Harish Gaggar
RoleBreak is an open benchmark designed to evaluate long‑horizon role‑playing robustness in spoken dialogue systems. It includes 310 character‑based and user‑centered roles, 6,688 human‑verified dialogue turns, and 11,743 fine‑grained evaluation criteria, with 1,856 turns specifically targeting expressive vocal emotion. The benchmark stresses role consistency, interaction quality, safety, and affect over extended conversations, and the authors evaluated nine system configurations across full‑duplex, omni‑modal, and cascaded ASR–LLM–TTS paradigms.
By Yuqi Wang, Fengyuan Liu, Haochen Luo, Zhiqi Yu, Qi Liu
The article surveys how large language models (LLMs) can be incorporated into networked control systems, cyber‑physical systems, and multi‑agent networks without violating stability and safety guarantees. It proposes treating the LLM as a slow supervisor that sets high‑level goals, while a fast, certified inner loop preserves physical stability. The survey maps LLM characteristics—such as inference latency, API failures, tokenization, and hallucinations—to classical control challenges and highlights the growing gap between model capability and formal safety assurances, calling for future research on stability proofs.
By Haiping Du, Linping Chan