npj Digital Medicine, Published online: 06 October 2026; doi:10.1038/s41746-026-03268-z

Machine learning classification of digital exclusion in ageing populations across five international cohorts

Nature Biomedical Engineering, Published online: 23 September 2026; doi:10.1038/s41551-026-01793-6

Contrary to expectations that unsaturation accelerates release, unsaturated phospholipid liposomes showed 3-fold higher loading and markedly reduced initial and sustained release due to multilamellar–multivesicular structure.

(Cell 188, 6953–6970.e1–e8; November 26, 2025)

arXiv:2609.13082v1 Announce Type: new Abstract: Agentic systems offer a promising way to automate embodied benchmark construction, but existing approaches typically cover isolated stages or remain specialized to predefined environments and task families. More importantly, multi-step construction produces dependent intermediate artifacts that are often passed downstream without artifact-specific verification, allowing local defects to propagate into the final benchmark. We present Embodied-BenchForge, an agentic framework that transforms user-specified evaluation intents into complete embodied benchmark artifacts. It formulates construction as Closed-Loop Benchmark Synthesis, integrating forward artifact synthesis with backward verification and repair. Skill-Orchestrated Artifact Synthesis composes typed and reusable skills into executable workflows, while an artifact dependency graph records intermediate outputs and their dependencies. Requirement-Guided Verification and Repair applies artifact-specific contracts throughout construction and uses provenance to trigger local re-execution or upstream rollback when verification fails. Embodied-BenchForge constructs six benchmarks covering diverse embodied scenarios in the Offline EQA Track, together with one interactive benchmark containing 220 executable tasks in the Interactive Embodied Track. Evaluations of representative MLLMs and embodied agents show that the benchmarks distinguish model capabilities in both observation-based understanding and closed-loop execution. Quality assessment and ablations validate benchmark quality and the effectiveness of verification and repair, while repair and skill-reuse analyses demonstrate efficient localized recovery and cross-benchmark reusability.

Background: Obstructive sleep apnea (OSA) is highly prevalent but remains substantially underdiagnosed. Polysomnography (PSG) is the reference standard, but its cost and limited availability constrain large-scale case identification. AI-based screening tools may support risk stratification and referral prioritization, but their diagnostic accuracy across apnea-hypopnea index (AHI) thresholds remains uncertain. Objective: This review aimed to systematically evaluate the diagnostic accuracy of AI-based OSA screening tools at AHI thresholds of ≥5, ≥15, and ≥30 events/hour, with emphasis on models using non-PSG–derived inputs. Methods: PubMed, Embase, Scopus, and Web of Science were searched for studies published from January 1, 2016, to May 3, 2026. Eligible studies included adults evaluated for suspected OSA or recruited from population-based cohorts, assessed AI-based models intended or interpretable for OSA screening, risk prediction, or screening-oriented severity classification, used PSG as the reference standard, and reported sufficient data to construct or reconstruct 2×2 contingency tables. Diagnostic accuracy was synthesized separately by AHI threshold and input source using bivariate random-effects models, with 95% CIs and prediction intervals (PIs). Risk of bias and certainty of evidence were assessed using QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies 2) and GRADE (Grading of Recommendations Assessment, Development, and Evaluation), respectively. Results: A total of 60 studies were included, of which 47 contributed data to the meta-analysis. At AHI thresholds of ≥5, ≥15, and ≥30 events/hour, pooled sensitivities were 0.94 (95% CI 0.92‐0.96; 95% PI 0.71‐0.99), 0.87 (95% CI 0.84‐0.89; 95% PI 0.66‐0.96), and 0.83 (95% CI 0.79‐0.87; 95% PI 0.61‐0.94), respectively; the corresponding specificities were 0.77 (95% CI 0.69‐0.84; 95% PI 0.30‐0.96), 0.81 (95% CI 0.75‐0.85; 95% PI 0.39‐0.96), and 0.91 (95% CI 0.87‐0.94; 95% PI 0.55‐0.99), respectively. The corresponding areas under the summary receiver operating characteristic curves were 0.943, 0.907, and 0.920. For non-PSG–derived tools, sensitivities were 0.92, 0.85, and 0.81, and specificities were 0.70, 0.74, and 0.85 at the 3 thresholds, respectively. For PSG-derived models, sensitivities were 0.96, 0.90, and 0.85, and specificities were 0.82, 0.88, and 0.96, respectively. Exploratory subgroup analyses suggested performance variation across selected study and model characteristics, including region, algorithmic framework, data source, and validation method. Conclusions: AI-based tools showed generally favorable screening performance for OSA across clinically relevant AHI thresholds, although wide PIs suggest variable performance across future comparable populations and settings. By synthesizing diagnostic accuracy across 3 AHI thresholds and distinguishing non-PSG–derived from PSG-derived models, this review extends previous broad or modality-specific reviews and offers a clinically interpretable, pathway-specific basis for linking model performance to intended use. The findings may clarify potential roles for non-PSG–derived tools in front-end screening and referral prioritization and for PSG-derived models in reduced-channel assessment and sleep-laboratory workflow support. Given substantial heterogeneity, limited external validation, and low or very low certainty of evidence, prospective validation is needed before routine implementation.

Cancer cachexia is a devastating metabolic syndrome for which there are no approved treatments. Li and colleagues developed an RNA-targeted therapy, which ameliorates cachectic symptoms, reduces inflammation, and extends survival in mouse cancer models. The study provides an approach for treating cancer cachexia and paves the road for clinical studies.

arXiv:2609.09578v1 Announce Type: new Abstract: Urban planning is a real-world spatial optimization problem that requires selecting feasible actions from large candidate spaces under practical objectives such as cost and service quality. Existing optimization and reinforcement learning methods are effective for fixed formulations, but often depend on task-specific representations and constraint handling. We propose \emph{CityPlanner}, a sandbox-agent framework for executable urban planning. CityPlanner introduces \emph{UrbanSandbox}, a unified file-based environment where agents inspect task files, generate plans, run evaluators, and revise decisions based on executable feedback. To make learning tractable, we further propose atomic-task reinforcement learning, which decomposes long sandbox trajectories into \emph{BuildPlan} for initial construction and \emph{ImprovePlan} for feedback-based refinement. Experiments on a real-world benchmark show that CityPlanner consistently outperforms heuristic, task-specific RL, and general LLM-agent baselines. Ablations verify the contributions of UrbanSandbox, atomic-task RL, and iterative deployment. We release the code and dataset at https://anonymous.4open.science/r/co-agent-C1C8

arXiv:2609.09646v1 Announce Type: new Abstract: Semantic-gradient-based prompt optimization (SGPO) improves agent harnesses using execution feedback, but its local update rule leaves the choice of edit scope and operation unresolved. We introduce RobustSGPO, which specifies the requested edit, constructs and checks the patch, and continues search from either the incumbent or retained snapshots. We evaluate permission scheduling, cumulative controls, and task-family transfer in the AgentX brainstorming workflow using 120 tasks, 95 runs, and 7,350 candidate attempts. Periodic $1\to2\to3$ scheduling exceeds fixed maximum permission by 0.28 test-score points. RobustSGPO increases completion on 30 held-out tasks from 60.0% to 80.0% and improves test quality from 3.77 to 4.14 under a 20-million-token budget. Category retention reduces source-task degradation after a shift, whereas random retention reaches a higher destination endpoint. Search-space control benefits quality through executable edits and alternative starting points, with measurable retention overhead.

arXiv:2609.09707v1 Announce Type: new Abstract: Supervised fine-tuning (SFT) applies a uniform cross-entropy loss to all target tokens, even though different tokens provide unequal learning signals for mathematical reasoning. This uniform treatment can over-sharpen already mastered tokens while amplifying learning pressure on uncertain, low-confidence tokens, leading to suboptimal training dynamics. We propose Trimmed Logit-Gap SFT (TrimSFT), a simple token-level reweighting method that scales the SFT loss according to the logit gap between the gold token and its strongest competitor. TrimSFT trims supervision away from both extremes: tokens already mastered (large logit gap) and tokens weakly supported by the current model (small or negative logit gap), concentrating learning within an intermediate logit-gap region between them. We instantiate this principle with a Gaussian weight centered at margin m with bandwidth {\tau}, requiring no reference model or additional forward pass. We evaluate TrimSFT on six base models from the Llama, Qwen, and DeepMath families across five mathematical reasoning benchmarks. TrimSFT consistently improves over standard SFT, achieving the best average performance on five out of six models, with gains of up to +26.9 points over SFT on MATH500. Further analyses show that the bandwidth {\tau} matters more than the exact margin location, and that half-trim variants that remove supervision pressure from only one side yield inferior trade-offs. A token-level logit-gap distribution analysis suggests that TrimSFT reshapes model confidence in a more balanced way than uniform SFT or monotonic reweighting methods. These results suggest that reasoning SFT can benefit from trimming both extremes rather than treating all tokens uniformly.

arXiv:2609.10092v1 Announce Type: new Abstract: Large language models (LLMs) increasingly act as research agents, yet their ability to track shifts in research attention is difficult to evaluate because reviews and research ideas lack uniquely verifiable outcomes. We introduce Research Attention Prediction (RAP), a rolling benchmark covering 278 AI/ML fields and 1,390 episodes. At each cut-off, an LLM agent searches a temporally restricted arXiv corpus and predicts the next six months' paper shares across eight frozen research directions. Search generally helps, but all four diagnostic models perform worse than an exact-count exponentially weighted moving average (EWMA) baseline in compositional accuracy. We identify two linked bottlenecks. Under cumulative-history access, State carry-forward outperforms direct Forecast for all four diagnostic models; frozen-evidence replay links a shared component of this reversal to Forecast-oriented policies retrieving a smaller share of recent evidence. Even with exact historical activity, future-specific updating remains limited, with only GPT-5.5 plus reopened Search slightly surpassing EWMA. Fine-tuning on realised outcomes improves Qwen3-4B's forecast Spearman correlation by 0.105 on held-out fields at later origins, with gains also on change-rich episodes.

arXiv:2609.10441v1 Announce Type: new Abstract: While Large Language Models (LLMs) have demonstrated impressive capabilities, they often struggle with extremely long contexts due to fixed context limits. To address this, sequential approaches like MemAgent extend the effective context by reading text in segments and iteratively updating a fixed-size memory. However, this sequential paradigm suffers from high latency and requires costly reinforcement learning (RL) training, which can lead to overfitting on specific datasets. To overcome these limitations, we propose ConvMem, a training-free, highly parallelizable framework that reformulates long-context reasoning as a hierarchical convolution. Inspired by CNNs, ConvMem treats an LLM prompted with a specific query as a convolutional kernel. This kernel summarizes text segments hierarchically, shortening the reasoning path from a linear chain into a logarithmic tree. Specifically, ConvMem integrates \textit{Configurable Strides} and \textit{Skip Connections} to ensure robust evidence capture and propagation, while employing \textit{Multi-Kernel Convolution} to decompose complex queries into disentangled semantic channels. This design not only mitigates error accumulation but also enables massive parallelization across both text segments and reasoning threads. Experiments on RULER-HotpotQA and RULER-2WikiMultiHopQA demonstrate that ConvMem outperforms training-free baselines and avoids the risk of overfitting to parametric priors often observed in RL-trained models on out-of-distribution tasks.

arXiv:2506.09084v2 Announce Type: replace-cross Abstract: Whole-page optimization (WPO) decides how search and recommendation results are surfaced to users, and large language models (LLMs) open a new route to it by treating page generation as sequence generation. Adapting LLMs to web-scale WPO, however, remains bottlenecked by the need for costly human annotations and by the mismatched granularity between page-level coherence and item-level placement. In this work we show that these two challenges are coupled: implicit user feedback alone suffices for alignment, provided the reward signal is decoupled into two complementary granularities. We propose PageLLM, a reward-based fine-tuning framework that (i) turns implicit feedback into four contrastive preference-pair families covering relevance, ranking, diversity, and redundancy, (ii) learns a coarse page-level reward and a fine item-level reward that captures engagement-sensitive position swaps, and (iii) combines both rewards in PPO-based RLHF over a pre-trained LLM. Extensive experiments on seven Amazon categories against eleven baselines show that neither reward alone is sufficient -- dropping the page-level or item-level signal reduces NDCG@100 by 17.8% and 15.2% respectively, whereas the joint reward improves NDCG@100 by up to 46.8%. Deployed in a 10M-user online A/B test, PageLLM raises GMV by 0.44% and click-through rate by 0.14%, confirming that multi-grained rewards from implicit feedback scale to production WPO. Code and data are available at an anonymized repository.

arXiv:2604.08213v2 Announce Type: replace-cross Abstract: High-quality source-target image pairs with precise editing instructions are essential for instruction-guided image editing, yet constructing such training triplets at scale remains costly. Recent pipelines often rely on vision-language models to synthesize editing instructions automatically, but we find that strong VLMs still struggle to describe visual transformations between image pairs. In particular, they exhibit three recurring failure modes: orientation inconsistency, viewpoint ambiguity, and missing fine-grained attributes. In a human evaluation on 400 image pairs, several open-source VLM baselines produce critical-error rates above 47\%, making many synthesized instructions unsuitable for downstream training. To address this, we propose EditCaption, a two-stage post-training pipeline for image editing instruction synthesis. First, we construct a 100K supervised fine-tuning dataset through GLM-based auto-captioning, EditScore filtering, and human refinement. Second, we collect 10K human-annotated preference pairs, where each rejected instruction is labeled with its primary error type and severity. Based on this dataset, we propose Hardness-Adaptive Error-Aware DPO (HAE-DPO), a task-adapted DPO objective that introduces an adaptive margin based on human-labeled severity, failure-mode type, and reference-model hardness. Experiments across three benchmarks demonstrate that our 235B model with SFT+HAE-DPO achieves state-of-the-art performance among open-source and closed models, scoring 4.720 on Eval-400, 4.672 on HQ-Edit, and 4.651 on ByteMorph-Bench -- surpassing Gemini-3-Pro on all three. Human evaluation confirms critical error rates drop from 47.75\% to 17.50\%, with correct rates improving from 41.75\% to 70.25\%, surpassing Gemini-3-Pro (66.00\%).

Oncogene, Published online: 19 May 2026; doi:10.1038/s41388-026-03823-8

NDRG2 orchestrates circadian clock stability to suppress tumorigenesis and potentiate oxaliplatin response in colorectal cancer

Oncogene, Published online: 18 April 2026; doi:10.1038/s41388-026-03781-1

Sirt1 sustains Sonic hedgehog signaling to promote medulloblastoma progression through regulating Gli3 processing

Nature, Published online: 15 April 2026; doi:10.1038/s41586-026-10416-8

A genome-to-genome association study identifies host and viral risk factors that interact to drive nasopharyngeal carcinoma endemicity in southern China.

Nature, Published online: 15 April 2026; doi:10.1038/s41586-026-10223-1

A composable neural network emulator is described for speeding up thermoelectric generator design, demonstrating the ability to predict generator performance with >99% accuracy while taking only 0.01% of the time compared with commercial finite-element solvers.

arXiv:2604.04383v1 Announce Type: new Abstract: Service system performance depends on how participants respond to design choices, but modeling these responses is hard due to the complexity of human behavior. We introduce an LLM-powered multi-agent simulation (LLM-MAS) framework for optimizing service operations. We pose the problem as stochastic optimization with decision-dependent uncertainty: design choices are embedded in prompts and shape the distribution of outcomes from interacting LLM-powered agents. By embedding key numerical information in prompts and extracting it from LLM-generated text, we model this uncertainty as a controlled Markov chain. We develop an on-trajectory learning algorithm that, on a single simulation run, simultaneously constructs zeroth-order gradient estimates and updates design parameters to optimize steady-state performance. We also incorporate variance reduction techniques. In a sustainable supply chain application, our method outperforms benchmarks, including blackbox optimization and using LLMs as numerical solvers or as role-playing system designers. A case study on optimal contest design with real behavioral data shows that LLM-MAS is both as a cost-effective evaluator of known designs and an exploratory tool that can uncover strong designs overlooked by traditional approaches.

arXiv:2604.04078v1 Announce Type: cross Abstract: Cardiac magnetic resonance (CMR) is a cornerstone for diagnosing cardiovascular disease. However, it remains underutilized due to complex, time-consuming interpretation across multi-sequences, phases, quantitative measures that heavily reliant on specialized expertise. Here, we present BAAI Cardiac Agent, a multimodal intelligent system designed for end-to-end CMR interpretation. The agent integrates specialized cardiac expert models to perform automated segmentation of cardiac structures, functional quantification, tissue characterization and disease diagnosis, and generates structured clinical reports within a unified workflow. Evaluated on CMR datasets from two hospitals (2413 patients) spanning 7-types of major cardiovascular diseases, the agent achieved an area under the receiver-operating-characteristic curve exceeding 0.93 internally and 0.81 externally. In the task of estimating left ventricular function indices, the results generated by this system for core parameters such as ejection fraction, stroke volume, and left ventricular mass are highly consistent with clinical reports, with Pearson correlation coefficients all exceeding 0.90. The agent outperformed state-of-the-art models in segmentation and diagnostic tasks, and generated clinical reports showing high concordance with expert radiologists (six readers across three experience levels). By dynamically orchestrating expert models for coordinated multimodal analysis, this agent framework enables accurate, efficient CMR interpretation and highlights its potentials for complex clinical imaging workflows. Code is available at https://github.com/plantain-herb/Cardiac-Agent.

arXiv:2604.02368v3 Announce Type: replace Abstract: As Large Language Models (LLMs) exhibit plateauing performance on conventional benchmarks, a pivotal challenge persists: evaluating their proficiency in complex, open-ended tasks characterizing genuine expert-level cognition. Existing frameworks suffer from narrow domain coverage, reliance on generalist tasks, or self-evaluation biases. To bridge this gap, we present XpertBench, a high-fidelity benchmark engineered to assess LLMs across authentic professional domains. XpertBench consists of 1,346 meticulously curated tasks across 80 categories, spanning finance, healthcare, legal services, education, and dual-track research (STEM and Humanities). These tasks are derived from over 1,000 submissions by domain experts--including researchers from elite institutions and practitioners with extensive clinical or industrial experience--ensuring superior ecological validity. Each task uses detailed rubrics with mostly 15-40 weighted checkpoints to assess professional rigor. To facilitate scalable yet human-aligned assessment, we introduce ShotJudge, a novel evaluation paradigm that employs LLM judges calibrated with expert few-shot exemplars to mitigate self-rewarding biases. Our empirical evaluation of state-of-the-art LLMs reveals a pronounced performance ceiling: even leading models achieve a peak success rate of only ~66%, with a mean score around 55%. Models also exhibit domain-specific divergence, showing non-overlapping strengths in quantitative reasoning versus linguistic synthesis.. These findings underscore a significant "expert-gap" in current AI systems and establish XpertBench as a critical instrument for navigating the transition from general-purpose assistants to specialized professional collaborators.