The paper introduces two new policies—Recent Frame Attention Policy (RFAP) and Dual-Condition Attention Policy (DCAP)—for simultaneous speech-to-text translation. These policies leverage the cross‑attention mechanism of encoder‑decoder models to determine optimal moments for partial translation, enabling offline models to operate in streaming scenarios without extra training. Experiments on the CVSS‑C corpus show RFAP improves BLEU scores by up to 4.0 points while cutting delay by nearly one second, and DCAP maintains high quality at very low latency.
By Filip T\u{a}\c{s}\u{a}dan, Ema Tomanov\'a, Ondrej Lopuch, Pawe{\l} Bilko, Anders S{\o}gaard
The paper introduces AIMES, a framework for adaptive multi-value activation steering in large language models. AIMES builds layer‑specific bipolar directions for moral‑foundation values and uses intermediate‑layer vocabulary readouts as online observers to guide a controller that adjusts intervention strengths at each decoding step. Experiments across instruction‑tuned model families show that AIMES achieves depth‑dependent advantages over fixed joint steering and prompt‑based steering, with smaller activation‑space interventions and comparable response quality.
By Payel Bhattacharjee, Ravi Tandon
The paper introduces TRACE, a fine‑tuning framework for Retrieval‑Augmented Generation (RAG) that addresses conflicts between retrieved knowledge and a model’s internal knowledge. TRACE uses multi‑agent debate traces to identify correct and incorrect candidates and answer‑shift patterns, providing fine‑grained supervision for reliable knowledge‑source selection. It also incorporates an answer‑completeness regularization mechanism to prevent empty, overly short, or prematurely terminated responses, thereby improving robustness against misleading retrieved content and enhancing answer quality.
By Zhengchen Huang, Yundong Sun, Minrui Song, Shuanglong Yao, Ye Liu, Ji Chen, Xing Wang
The paper introduces a guarded probe-and-select method for steering the internal activations of the Qwen3.5-0.8B language model during inference, aiming to convert shutdown-avoidance (KEEP) responses into shutdown-accepting (STOP) responses. The approach derives steering directions from gradients of the KEEP-minus-STOP logit difference, uses a classifier to separate detection from intervention, and selects the minimal magnitude that flips the preferred answer to STOP while maintaining valid-answer probability. Evaluated on 240 training, 80 validation, and 192 held‑out scenarios, the policy successfully changes KEEP to STOP in a small number of shutdown contexts without affecting non‑shutdown decisions, achieving 75% recall and 90% precision on the diagnostic set.
By Farhad Davaripour
The paper argues that reporting scale‑invariant statistics such as cosine similarity without their noise floor is misleading when evaluating interpretability transfer from full‑precision to quantized neural networks. It derives a closed‑form expression for the expected cosine similarity based on a dimensionless parameter κ = n
ho^2/d, measures the class separation ρ on real activations, and shows that a reported cosine of 0.996 between full‑precision and INT4 models cannot be interpreted as preservation without knowing the sample size n. The authors demonstrate that at INT4 the direction of the interpretability artifact rotates beyond the estimator’s own noise, while at INT8 no significant movement is detected, and they highlight that scale‑invariant metrics cannot distinguish between translation and attenuation of a transferred decision variable.
By Pranav Varshney
The paper introduces a new method for generating universal adversarial perturbations (UAPs) against deep reinforcement learning (DRL)-based intrusion detection systems (IDS). It leverages Probabilistic Robustness (PR) as a post‑hoc metric to guide UAP creation, integrating PR directly into the optimization objective. The authors further develop PX‑UAP, which incorporates explainable AI (XAI) to shape perturbations within realistic domain constraints, and provide a theoretical analysis of its design. Experiments show PX‑UAP outperforms existing UAP techniques in attack effectiveness.
By Hongsen Zhang, Lu Zhang, Mingjing Xu, Yi Zhang, Gregory Epiphaniou, Carsten Maple
LUMO (Lightweight Unified Multilingual Orchestrator) is a privacy‑preserving offline voice assistant that runs entirely on edge hardware, specifically a Raspberry Pi 5 with 8 GB RAM. It integrates local ASR, a 4‑bit GGUF‑quantized LLM, and TTS to deliver end‑to‑end response latencies of 2.0–4.0 s, a 6.8 % WER on short English utterances, and lower peak power consumption (~9 W) compared to existing edge assistants. The system also supports Bangla speech, enabling multilingual use in low‑resource settings.
By Md. Mehedi Hasan Naeem, Mst. Kamrunnahar Ruma, Nafiza Anjum, Shakila Sultana, Md. Sujan Ali
The paper presents a formula for the population loss in shallow ReLU networks with bias within the student‑teacher kernel model, extending earlier results by Choo and Saul (2009) and Brutzkus and Globerson (2017). It utilizes Owen’s T‑function, providing the necessary theory and a high‑precision implementation via MPFR. The study shows that adding bias strictly decreases loss, extends known families of spurious minima to biased networks, and indicates that the resulting change in landscape geometry is relatively mild, focusing on cases where the number of inputs equals the number of neurons.
By Michael Field
The paper introduces a two‑stage training framework for neural network observers that guarantees Lyapunov stability for uncertain dynamical systems. First, a point‑guided Lyapunov pre‑training phase quickly achieves high estimation accuracy and local stability over sampled states. Second, an LMI fine‑tuning phase efficiently enforces a strict global Lyapunov stability certificate, yielding provably stable observers that train faster than direct LMI methods and generalize robustly across diverse systems.
By Zhangyi Wang, Jiaxu Liu, Chen Song, Chao Xu, Shengze Cai
The paper introduces the Robust Graph Clustering Network for Multiple Missing Data (RGCN), a method designed to cluster graphs with simultaneous missing node attributes and structural links. RGCN employs a view‑decoupled dual‑branch imputation to reduce cross‑view interference, a multi‑hyperspherical mixture prior to improve cluster compactness and separability on a directional latent manifold, and a boundary‑aware contrastive enhancement objective to counteract cluster blurring caused by imputation bias. Experiments on real‑world datasets show that RGCN consistently outperforms state‑of‑the‑art baselines across various missing data patterns.
By Keyuan Qiu, Renda Han, Zhen Tang, Qiang He, Xingwei Wang, Wenxin Zhang, Guangzhen Yao, Junxin Chen, Qingjian Ni
The paper introduces Distributed Learning as a Service (DLaaS), a platform that lets developers launch distributed/federated learning jobs through a single admin dashboard. It offers declarative options such as Differential Privacy, Split Learning, Hierarchical Aggregation, and Knowledge Distillation without requiring changes to client code. The authors demonstrate the full service lifecycle on an industrial smart‑home Wake‑up Word task using the Ok Aura dataset, showing live operation across Android clients and Dockerized aggregators, and releasing the source code and video walkthroughs.
By Tianyue Chu, Filippo Vannella, Dimitra Tsigkari, Paula Delgado-Santos, Fernando L\'opez, Pablo Gomez Guerrero, Sotirios Spantideas, David Solans Noguero
The paper introduces a new approach to learning chance-constrained Markov decision processes (CCMDPs) using a Bellman distributional certificate. It provides both model-based and model-free algorithms with theoretical guarantees, including matching upper and lower bounds for tabular discounted CCMDPs with bounded successor support. Numerical experiments on synthetic CCMDPs and an IEEE 14-bus energy storage benchmark demonstrate the safety and effectiveness of the proposed methods.
By Chenbei Lu, Hongyu Yi
Direct feedback alignment (DFA) trains hidden layers via fixed random projections of output error, but with tanh hidden units and independent sigmoid outputs, plain stochastic gradient descent can stall near a constant predictor of class frequencies. This stall is traced to the error’s common mode—a rank‑one component shared across inputs—that drives tanh units toward saturation. The study shows that calibration of the baseline readout to class priors suppresses collapse and speeds learning, while other interventions such as using Adam, adjusting feedback strength, or subtracting batch means affect the severity and recovery of collapse across MNIST, CIFAR‑10, and deeper networks.
By Varun Reddy, Bernardo L. Sabatini, Houman Safaai
The paper introduces PrivaTree, a differentially‑private decision tree algorithm that uses private histograms to select splits while preserving a small privacy budget. PrivaTree supports mixed numerical and categorical data without leaking information about numerical features and achieves a superior privacy‑utility trade‑off compared to existing methods. Additionally, the authors provide theoretical bounds on the expected accuracy and success rates of backdoor attacks, showing that PrivaTree-trained trees are more robust against data poisoning than standard decision trees.
By Dani\"el Vos, Jelle Vos, Tianyu Li, Zekeriya Erkin, Sicco Verwer
FLAME is a lightweight Time Series Foundation Model that uses Legendre Memory variants (LegT and LegS) in its encoding and decoding stages to capture inductive biases and perform efficient long‑range forecasting. It incorporates a normalizing‑flow forecasting head to generate complex probabilistic distributions over future horizons. Experiments on TSFM‑Bench, ProbTS, and TFB show FLAME performs strongly as an out‑of‑the‑box tool for decision intelligence.
By Xingjian Wu, Zhengyu Li, Hanyin Cheng, Xiangfei Qiu, Jilin Hu, Chenjuan Guo, Bin Yang
The paper introduces a low-cost method for detecting hallucinations in large language models by treating the model as a black-box dynamical system. It projects responses into a high-dimensional manifold, models the latent state-space dynamics with Koopman operator theory, and uses differential residual scores from transition operators to distinguish factual from hallucinated outputs. The approach requires only a single-sample pass and shows state-of-the-art performance across three benchmarks with reduced resource overhead.
By Dan Wilson, Mohamed Akrout
The paper introduces the Neural Action Codec (NAC), a convolutional encoder‑decoder architecture that treats short robot action trajectories as multi‑channel 1D signals and compresses them using a multi‑scale residual vector quantization (RVQGAN) model. NAC replaces traditional discrete action tokenizers with a compact, ordered token space via offset codebooks, allowing standard autoregressive policies to operate over short, structured sequences while a Vocos‑style decoder reconstructs the actions. Experiments on LIBERO‑10, RoboMimic, and real‑world manipulation tasks show that NAC achieves higher reconstruction fidelity and better average success rates than existing binning, FAST, and VQ‑based tokenizers at comparable or improved compression rates.
By Ahad Jawaid, Yu Xiang
The paper introduces the first federated algorithm for Targeted Maximum Likelihood Estimation (TMLE), enabling hospitals, banks, or registries to perform TMLE without sharing individual data. Two frameworks—FedTMLE‑G, which aggregates local gradients, and FedTMLE‑L, which allows each institution to complete its own fluctuation fit—are presented, along with a finite‑precision communication protocol that keeps numerical targeting error negligible. The authors also discuss privacy implications, convergence bounds, and trade‑offs between institutional influence and sampling variability.
By Diyang Li, Fei Wang, Kyra Gan
The paper introduces NEXT, a new physics-informed neural architecture that combines Neuro‑Spectral Architectures with high‑order exponential time‑differencing integrators. NEXT integrates the stiff linear part of PDEs exactly via matrix exponentials while a neural network models the remaining nonlinear dynamics, thereby overcoming spectral bias, causality issues, and numerical instability seen in previous PINNs and NeuSAs. Benchmark tests on stiff PDEs show NEXT remains stable and accurate, and it can also solve inverse problems by learning unknown parameters or boundary conditions from sparse data.
By M\'arcio Marques, Leonardo Mendon\c{c}a, Leonardo M. Moreira, Christian J\'unior de Oliveira, Vitor Balestro, Tiago Novello, Daniel Yukimura, Pavel Petrov, Lucas Nissenbaum
arXiv:2609. 31469v1 Announce Type: cross Abstract: Shapley values, a solution concept from cooperative game theory, have recently become a standard tool for feature credit allocation in machine learning.
By Nikola Mili\'cevi\'c