The paper introduces Personalized Federated Hierarchical Gaussian Processes (pFedHGP), a method for probabilistic regression and classification on data distributed across heterogeneous clients. Each client’s latent function is split into a shared global component, a client‑specific deviation that shares the global kernel, and a flexible local residual. Using sparse inducing‑variable approximations and federated variational inference, raw data remain local while the server exchanges only low‑dimensional statistics, enabling full predictive distributions for uncertainty‑aware decisions. In experiments, pFedHGP achieves perfect fault classification in press tonnage monitoring with only 13.77% of labeled cycles and accurately recovers geographic zones in federated air‑quality modeling without centralizing station‑level time series.
By Xianjian Xie, Hao Yan
The paper compares three recurrent-depth transformer architectures—Block Universal Transformer (BUT), CoTFormer, and a new LSTM Universal Transformer (LSTM‑UT)—on cellular automata and delayed‑recall tasks. BUT consistently extrapolates to unseen depths better than CoTFormer, though its accuracy eventually falls. LSTM‑UT, which uses bounded gated memory, outperforms both baselines in depth extrapolation and delayed recall, demonstrating the value of gated memory for repeated computation and retrieval.
By Aras Kavuncu
The paper demonstrates that adding stochastic noise to a consolidation rule can improve a neural network’s retention of earlier tasks up to an optimal level, after which performance degrades, forming an inverted‑U relationship. Through simulations on related‑task continual‑learning benchmarks, the authors isolate the conditions that produce this optimum, showing it requires coherent restoration toward consolidated weights and is linked to the noise variance. The study further maps the scope of the effect, noting it depends on shared task structure and diminishes with more tasks, while a single‑seed hardware demonstration is referenced elsewhere.
By Gunner Levi Howe
The paper investigates why on‑policy distillation (OPD) can produce excessively long student responses, attributing this to a termination‑token mismatch between base students and post‑trained teachers. Experiments on Qwen3, Llama, and Gemma show that differing stopping probabilities for identical EOS token sets suppress the student’s preferred termination and fail to transfer the teacher’s alternative. Aligning decoding stopping sets alone is insufficient; treating functionally equivalent EOS tokens as a shared semantic stopping action reduces length inflation, though late‑stage inflation persists beyond termination alignment.
By Yuxiao Yang, Tianrun Yu, Shangzhe Li, Kaixiang Zhao, Xuchao Zhang, Chetan Bansal, Huaxiu Yao, Taylor W. Killian, Weitong Zhang
NS3Learn is a closed‑form model that captures realistic 5G NR sidelink Mode‑2 reception losses—such as half‑duplex loss, scheduling collisions, receiver capture, and decoding—by fitting 10.5 million labeled outcomes from ns‑3 5G‑LENA traces. The model achieves a mean absolute deviation of 0.06 in per‑instant delivery compared to ns‑3, outperforming alternative models, and its parameters transfer with minimal error to new intersections. Using NS3Learn in traffic‑network simulations reverses traffic speed trends and more than doubles predicted hard‑braking events, demonstrating its impact on safety assessments.
By Rasheed Bello, Arthur Mukwaya, Gurcan Comert, Varghese Vaidyan, Vijay Bendigeri, Anthony Dontoh, Jagruti Sahoo, Judith Mwakalonge
The paper introduces L2C‑TFM, a reinforcement‑learning framework for cleaning tabular data before feeding it to Tabular Foundation Models (TFMs). It proposes a model‑aware reward that regularizes the Wasserstein distance between cleaned and dirty data, aiming to preserve distributional stability. Experiments on ten OpenML datasets show that while some reward designs fail, the model‑aware reward performs comparably to a random‑forest baseline and improves minority‑class macro‑F1 under class imbalance, and a policy trained on one dataset can transfer to others.
By Laure Berti-Equille
arXiv:2606. 01527v2 Announce Type: replace Abstract: Machine unlearning is motivated by legal and user-facing requirements to remove the influence of individuals' data from trained models, such as the right to be forgotten.
By Matthew Regehr, Gautam Kamath, Andrew Lowy
VGAS: Variance-Reduced Guidance and Adaptive Selection for Training-Free Reward Alignment in Discrete Diffusion proposes a new inference-time framework that improves steering of frozen masked discrete diffusion models. By reducing the variance of guidance estimates, applying reward tilting to clean-token logits, and adapting the selection temperature at each step, VGAS addresses three default choices in existing pipelines. Experiments on regulatory DNA, protein, and small-molecule benchmarks show that VGAS achieves the best training-free reward performance and matches or surpasses reward-fine-tuned generators.
By Kwanyoung Kim
Compass‑v3 is a 245B‑parameter Mixture‑of‑Experts language model tailored for Southeast Asian e‑commerce, featuring 71B active parameters per token and hardware‑efficient expert parallelism. It is trained on 12 trillion multilingual tokens and synthetic e‑commerce instructions, and incorporates Optimal‑Transport Direct Preference Optimization to improve instruction adherence. Benchmarks show it outperforms GPT‑4, DeepSeek‑V3.1, and Qwen3‑235B, and it is already deployed at scale on Shopee, handling over 70% of the platform’s LLM traffic.
By Sophia Maria
CORTEX is a novel framework that transforms web‑scale corpus construction from flat document filtering into structured knowledge organization using an Ontological Corpus Graph (OCG). The OCG comprises a quality‑refined content layer, a lightweight ontology layer that evolves via LLMs, and a cross‑domain alignment layer that supports arbitrary taxonomic resolution. Experiments demonstrate CORTEX’s effectiveness, and the authors release a 24.14 B‑token refined corpus, its OCG, and a cross‑domain benchmark called CortexBench for evaluating large language models.
By Chengtao Gan, Xiaoke Guo, Yushan Zhu, Zhaoyan Gong, Zhiqiang Liu, Songze Li, Huajun Chen, Wen Zhang
The paper introduces a new task called visual scene negative captioning, which aims to describe what should be present in an image but is actually absent, a capability crucial for safety-critical applications. It proposes the CRCD framework, which uses counterfactual reconstruction and contrastive decoding to overcome affirmation bias, limited mental filling, and representation bias. CRCD employs a dual-branch architecture for amodal completion and functional association, along with multi-condition representation learning, to generate accurate negative captions and sets a high-performance baseline for this emerging task.
By Zhiyun Jiang, Hanyong Wang, Binbin Liang, Yu Xie, Menglong Yang, Wei Li
The paper investigates whether large language models (LLMs) can internally detect harmful content, bypassing external guardrails that add latency and computational cost. By extracting activations from LLaMA‑3.1‑8B and training lightweight MLP probes, the authors achieve high F1 scores (99%, 83%, and 84%) on WildJailbreak, Beavertails, and AEGIS 2.0 benchmarks, rivaling much larger guard models while reducing overhead. This suggests that internal state monitoring can provide efficient safety checks for resource‑constrained, time‑critical deployments.
By Alizishaan Khatri, Chiquita Prabhu, Omkar Neogi
The paper introduces the Generic Multi-Agent Trading System (GMATS), a framework for studying how large language model (LLM) based trading stacks react to black-box, input-only attacks that inject plausible social‑media content. It defines contagion metrics—belief‑shift scores at analyst and coordinator layers and attack‑clean deltas on backtest metrics—to trace the spread of adversarial signals. Experiments on a safe offline benchmark show that even simple attackers can significantly degrade risk‑return profiles, while certain multi‑agent topologies and coordinator prompts can mitigate these effects.
By Qi Rong Sua, Junhao Dong, Nguyen Duc Thai, Yuqing Wen, Cheston Tan, Yew-Soon Ong
Code-as-Auditor is an LLM-based framework that transforms regulatory information into formal checklists and executable decision trees, encoding rules as interpretable code. During inference, the model expands each checklist item into factual and counterfactual questions, guiding reasoning over case-specific evidence and potential violations. This pipeline moves from evidence identification to rule application and final decision-making, with a self‑verification loop that enhances logical consistency and traceability, leading to more accurate and evidence‑backed compliance evaluations in privacy and data protection scenarios.
By Jisoo Kim, Taeyoon Kwack, Jinwoo Jang, Woo Kyung Kim, Honguk Woo
AUDITPLAN introduces a plan-then-answer method for safety alignment in language models, where the model first generates a compact structured safety plan before responding. The plan includes a threat label, intended action, and explicit constraints, allowing machine‑checkable auditing while remaining hidden from end users. Training combines supervised fine‑tuning with reinforcement learning using the FAITHGATE reward, which only rewards correct plans, thereby reducing unsafe shortcuts and improving robustness across Qwen model variants.
By Sai Sri Pushpa Jampani, Kshitij Mishra, Asif Ekbal
The paper investigates what gradient similarity measures in data attribution for large language models. By independently varying task and answer format in supervised fine‑tuning benchmarks, the authors show that gradient alignment is driven by answer format rather than task semantics, with strong alignment for shared formats and none for differing formats. This pattern persists across training stages, model sizes, and families, and is evident in the selections of the LESS data‑selection method, which over‑represents its own answer format.
By Sunwoo Kim, Seokwon Jung, Sohyung Kim, Seong Joon Oh, Alice Oh
ZeroHAT is a new framework for generating synthetic human activity traces (HATs) in a target region without any real data from that region. It transfers behavioral patterns learned from real HATs in source regions and adapts them using publicly available contextual information about the target region. The system includes a consistency-aware intent extractor, a cross-region behavioral cloning module, and a behavior-conditioned activity realization module, and it outperforms the strongest baseline by 4.5–6.4× in downstream utility and improves fidelity by 15.6–40.8% across ten cities.
By Rongchao Xu, Dahai Yu, Lin Jiang, Guang Wang
The paper introduces the adversarial water-filling (AWF) framework for competitive resource allocation in frequency and space, particularly targeting multi-operator low Earth orbit satellite spectrum sharing. It presents theoretical results, algorithms, and a domain‑specific wireless foundation model that uses permutation‑invariant channel representations, a constraint‑aware graph neural network, and learned projected extragradient iterations to approximate stationary solutions of the minimax problem. Experiments show the model generalizes to unseen problem sizes and achieves over an order‑of‑magnitude speedup compared to Mirror‑Prox while maintaining comparable solution quality.
By Xindi Tong, Chee Wei Tan, H. Vincent Poor
SAFARI is the first industrial benchmark for evaluating large language models (LLMs) in automotive hazard analysis and risk assessment (HARA) under ISO 26262. It comprises 3,000 de‑identified HARA cases and tests two tasks: open‑ended hazard generation and standards‑grounded risk classification, using a novel reference‑anchored LLM‑as‑a‑judge protocol. Experiments with nine state‑of‑the‑art LLMs show that while hazard narratives are often plausible, risk classification remains weak (best ASIL macro‑F1 = 0.261), with errors mainly due to missing scenario context and misjudged controllability.
"whyItMatters":"The benchmark highlights the current limitations of LLMs in safety‑critical engineering workflows, guiding future research and expert oversight in automotive safety analysis."
By Chenxi Wu, Zimu Wang, Haiyang Zhang, Wei Wang, Zhijie Xu
The paper evaluates bias in phoneme-based automatic speech recognition (ASR) systems, focusing on WhisperIPA and ZIPA, which produce International Phonetic Alphabet (IPA) transcriptions. Using multilingual speech corpora and demographically annotated English datasets, the authors compare model-generated IPA against grapheme-to-phoneme (G2P) outputs with both standard phoneme error rate (PER) and a new Soft PER metric that allows linguistically similar substitutions. The study finds persistent disparities across language, gender, accent, ethnicity, and age, even when accounting for acceptable phonemic variation.
By Maneesha Rani Saha, Catherine Bao, Neal Patwari