arXiv:2601.17310v1 Announce Type: new Abstract: Simulation is a powerful tool for exploring uncertainty. Its potential in clinical medicine is transformative and includes personalized treatment planning and virtual clinical trials. However, simulating patient trajectories is challenging because of complex biological and sociocultural influences. Here, we show that real-world clinical records can be leveraged to empirically model patient timelines. We developed a generative simulator model that takes a patient's history as input and synthesizes fine-grained, realistic future trajectories. The model was pretrained on more than 200 million clinical records. It produced high-fidelity future timelines, closely matching event occurrence rates, laboratory test results, and temporal dynamics in real patient future data. It also accurately estimated future event probabilities, with observed-to-expected ratios consistently near 1.0 across diverse outcomes and time horizons. Our results reveal the untapped value of real-world data in electronic health records and introduce a scalable framework for in silico modeling of clinical care.

arXiv:2601.18641v1 Announce Type: cross Abstract: Speech remains one of the most visible yet overlooked vectors of inclusion and exclusion in contemporary society. While fluency is often equated with credibility and competence, individuals with atypical speech patterns are routinely marginalized. Given the current state of the debate, this article focuses on the structural biases that shape perceptions of atypical speech and are now being encoded into artificial intelligence. Automated speech recognition (ASR) systems and voice interfaces, trained predominantly on standardized speech, routinely fail to recognize or respond to diverse voices, compounding digital exclusion. As AI technologies increasingly mediate access to opportunity, the study calls for inclusive technological design, anti-bias training to minimize the impact of discriminatory algorithmic decisions, and enforceable policy reform that explicitly recognize speech diversity as a matter of equity, not merely accessibility. Drawing on interdisciplinary research, the article advocates for a cultural and institutional shift in how we value voice, urging co-created solutions that elevate the rights, representation, and realities of atypical speakers in the digital age. Ultimately, the article reframes speech inclusion as a matter of equity (not accommodation) and advocates for co-created AI systems that reflect the full spectrum of human voices.

arXiv:2409.07314v2 Announce Type: replace-cross Abstract: While Large Language Models (LLMs) achieve superhuman performance on standardized medical licensing exams, these static benchmarks have become saturated and increasingly disconnected from the functional requirements of clinical workflows. To bridge the gap between theoretical capability and verified utility, we introduce MEDIC, a comprehensive evaluation framework establishing leading indicators across various clinical dimensions. Beyond standard question-answering, we assess operational capabilities using deterministic execution protocols and a novel Cross-Examination Framework (CEF), which quantifies information fidelity and hallucination rates without reliance on reference texts. Our evaluation across a heterogeneous task suite exposes critical performance trade-offs: we identify a significant knowledge-execution gap, where proficiency in static retrieval does not predict success in operational tasks such as clinical calculation or SQL generation. Furthermore, we observe a divergence between passive safety (refusal) and active safety (error detection), revealing that models fine-tuned for high refusal rates often fail to reliably audit clinical documentation for factual accuracy. These findings demonstrate that no single architecture dominates across all dimensions, highlighting the necessity of a portfolio approach to clinical model deployment. As part of this investigation, we released a public leaderboard on Hugging Face.\footnote{https://huggingface.co/spaces/m42-health/MEDIC-Benchmark}

arXiv:2502.16944v2 Announce Type: replace-cross Abstract: In this paper, we explore how directly pretraining a value model simplifies and stabilizes reinforcement learning from human feedback (RLHF). In reinforcement learning, value estimation is the key to policy optimization, distinct from reward supervision. The value function predicts the \emph{return-to-go} of a partial answer, that is, how promising the partial answer is if it were continued to completion. In RLHF, however, the standard pipeline first pretrains a reward model and then learns a value function online, even though no new reward signals are available once preference data is collected. This makes critic learning redundant, as the process of training a reward model and then deriving a value model is informationally equivalent to directly pretraining a value model. Importantly, this requires no additional supervision, and our value model is trained on exactly the same data used for reward modeling. Building on this insight, we introduce \emph{Decoupled Value Policy Optimization} (DVPO), a framework that pretrains a \emph{Global Value Model} (GVM) offline and freezes it as a universal critic for policy learning. The GVM provides stable, fine-grained credit assignment without critic drift or trajectory sampling. Experiments across MT-Bench, Alpaca-Eval, and Arena-Hard demonstrate that DVPO matches or surpasses state-of-the-art RLHF methods. These results highlight RLHF can be reframed as policy-only optimization guided by a single pretrained value model.

arXiv:2506.11238v2 Announce Type: replace-cross Abstract: Introduction: Premature Ventricular Contractions (PVCs) are common cardiac arrhythmias originating from the ventricles. Accurate detection remains challenging due to variability in electrocardiogram (ECG) waveforms caused by differences in lead placement, recording conditions, and population demographics. Methods: We developed uPVC-Net, a universal deep learning model to detect PVCs from any single-lead ECG recordings. The model is developed on four independent ECG datasets comprising a total of 8.3 million beats collected from Holter monitors and a modern wearable ECG patch. uPVC-Net employs a custom architecture and a multi-source, multi-lead training strategy. For each experiment, one dataset is held out to evaluate out-of-distribution (OOD) generalization. Results: uPVC-Net achieved an AUC between 97.8% and 99.1% on the held-out datasets. Notably, performance on wearable single-lead ECG data reached an AUC of 99.1%. Conclusion: uPVC-Net exhibits strong generalization across diverse lead configurations and populations, highlighting its potential for robust, real-world clinical deployment.

arXiv:2511.12869v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) have benefited enormously from scaling, yet these gains are bounded by five fundamental limitations: (1) hallucination, (2) context compression, (3) reasoning degradation, (4) retrieval fragility, and (5) multimodal misalignment. While existing surveys describe these phenomena empirically, they lack a rigorous theoretical synthesis connecting them to the foundational limits of computation, information, and learning. This work closes that gap by presenting a unified, proof-informed framework that formalizes the innate theoretical ceilings of LLM scaling. First, computability and uncomputability imply an irreducible residue of error: for any computably enumerable model family, diagonalization guarantees inputs on which some model must fail, and undecidable queries (e.g., halting-style tasks) induce infinite failure sets for all computable predictors. Second, information-theoretic and statistical constraints bound attainable accuracy even on decidable tasks, finite description length enforces compression error, and long-tail factual knowledge requires prohibitive sample complexity. Third, geometric and computational effects compress long contexts far below their nominal size due to positional under-training, encoding attenuation, and softmax crowding. We further show how likelihood-based training favors pattern completion over inference, how retrieval under token limits suffers from semantic drift and coupling noise, and how multimodal scaling inherits shallow cross-modal alignment. Across sections, we pair theorems and empirical evidence to outline where scaling helps, where it saturates, and where it cannot progress, providing both theoretical foundations and practical mitigation paths like bounded-oracle retrieval, positional curricula, and sparse or hierarchical attention.

arXiv:2601.15333v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) possess strong representation and reasoning capabilities, but their application to structure-based drug design (SBDD) is limited by insufficient understanding of protein structures and unpredictable molecular generation. To address these challenges, we propose Exploration-Augmented Latent Inference for LLMs (ELILLM), a framework that reinterprets the LLM generation process as an encoding, latent space exploration, and decoding workflow. ELILLM explicitly explores portions of the design problem beyond the model's current knowledge while using a decoding module to handle familiar regions, generating chemically valid and synthetically reasonable molecules. In our implementation, Bayesian optimization guides the systematic exploration of latent embeddings, and a position-aware surrogate model efficiently predicts binding affinity distributions to inform the search. Knowledge-guided decoding further reduces randomness and effectively imposes chemical validity constraints. We demonstrate ELILLM on the CrossDocked2020 benchmark, showing strong controlled exploration and high binding affinity scores compared with seven baseline methods. These results demonstrate that ELILLM can effectively enhance LLMs capabilities for SBDD.

npj Digital Medicine, Published online: 26 January 2026; doi:10.1038/s41746-025-02268-9

Multimodal digital biopsy for preoperative prediction of occult peritoneal metastasis in gastric cancer

Background: Home health aides and attendants (HHAs) provide in-home care to the growing population of older adults who want to age in place. Despite their vital role in patient care, HHAs are an underserved and vulnerable population of health care professionals who often experience poor health themselves. Activity tracking devices offer a promising way to improve HHAs’ health-related awareness and promote health behavior change, particularly regarding physical activity and sleep quality, 2 areas in which the workforce struggles. Objective: This study aimed to understand how feasible it is for HHAs to use activity tracking devices and assess their perceptions of such devices for improving their health. Specifically, we conducted (1) a field study to assess the use, feasibility, and acceptability of these devices among HHAs and (2) a qualitative study to understand HHAs’ perspectives on and reactions to activity trackers on and off the job. Methods: We partnered with the 1199 Service Employees International Union Training and Employment Fund to conduct a field study with home care agency–employed HHAs working in New York City, New York. Participants wore activity tracking devices for 4 weeks that collected data on physical activity and sleep. The HHAs were subsequently interviewed on their experiences with and attitudes toward the devices and asked to reflect on personalized visualizations of their data to prompt them to think aloud. Quantitative data were analyzed using descriptive statistics. Qualitative data were analyzed using grounded theory. Results: A total of 17 HHAs participated; their mean age was 48.7 (SD 12.2) years, 15 (88%) were women, 11 (65%) identified as Black, 5 (29%) identified as Hispanic or Latinx, and they had worked as HHAs for a mean of 11.7 (SD 7.5) years. In total, 94% (n=16) of the HHAs wore their activity trackers for the full 28-day study period. Participants took a mean of 10,230 (SD 3586) daily steps during the study period and slept for a mean of 6.27 (SD 0.58) hours per night. Overall, 4 key themes emerged: (1) activity tracking devices enhanced participants’ health awareness by providing empirical data for self-reflection; (2) this increased awareness led to positive behavior changes, including setting and achieving health-related goals; (3) HHAs believed that these devices could improve not only their own health but also that of their patients through positive behavior changes; and (4) despite this optimism, participants emphasized that their ability to modify sleep and activity patterns was constrained by social and occupational determinants, with sleep improvements being particularly challenging. Conclusions: Our findings suggest that appropriately designed personal tracking interventions could offer a promising approach to supporting positive health-related changes in this historically overlooked workforce, potentially improving their well-being and the quality of care they provide to their patients.

This virtual panel brings together engineers, architects, and technical leaders to explore how AI is changing the landscape of software development. Practitioners share their insights on successes and failures when AI is incorporated into daily workflows, emphasizing the significance of context, validation, and cultural adaptation in making AI a sustainable element of modern engineering practices.

By Arthur Casals, Mariia Bulycheva, May Walter, Phil Calçado, Andreas Kollegger

arXiv:2601.16414v1 Announce Type: cross Abstract: Difficulty replicating baselines, high computational costs, and required domain expertise create persistent barriers to clinical AI research. To address these challenges, we introduce PyHealth 2.0, an enhanced clinical deep learning toolkit that enables predictive modeling in as few as 7 lines of code. PyHealth 2.0 offers three key contributions: (1) a comprehensive toolkit addressing reproducibility and compatibility challenges by unifying 15+ datasets, 20+ clinical tasks, 25+ models, 5+ interpretability methods, and uncertainty quantification including conformal prediction within a single framework that supports diverse clinical data modalities - signals, imaging, and electronic health records - with translation of 5+ medical coding standards; (2) accessibility-focused design accommodating multimodal data and diverse computational resources with up to 39x faster processing and 20x lower memory usage, enabling work from 16GB laptops to production systems; and (3) an active open-source community of 400+ members lowering domain expertise barriers through extensive documentation, reproducible research contributions, and collaborations with academic health systems and industry partners, including multi-language support via RHealth. PyHealth 2.0 establishes an open-source foundation and community advancing accessible, reproducible healthcare AI. Available at pip install pyhealth.

arXiv:2601.16478v1 Announce Type: cross Abstract: With the rapid growth of scientific literature, scientific question answering (SciQA) has become increasingly critical for exploring and utilizing scientific knowledge. Retrieval-Augmented Generation (RAG) enhances LLMs by incorporating knowledge from external sources, thereby providing credible evidence for scientific question answering. But existing retrieval and reranking methods remain vulnerable to passages that are semantically similar but logically irrelevant, often reducing factual reliability and amplifying hallucinations.To address this challenge, we propose a Deep Evidence Reranking Agent (DeepEra) that integrates step-by-step reasoning, enabling more precise evaluation of candidate passages beyond surface-level semantics. To support systematic evaluation, we construct SciRAG-SSLI (Scientific RAG - Semantically Similar but Logically Irrelevant), a large-scale dataset comprising about 300K SciQA instances across 10 subjects, constructed from 10M scientific corpus. The dataset combines naturally retrieved contexts with systematically generated distractors to test logical robustness and factual grounding. Comprehensive evaluations confirm that our approach achieves superior retrieval performance compared to leading rerankers. To our knowledge, this work is the first to comprehensively study and empirically validate innegligible SSLI issues in two-stage RAG frameworks.

arXiv:2601.16895v1 Announce Type: cross Abstract: Surgery is a highly complex process, and artificial intelligence has emerged as a transformative force in supporting surgical guidance and decision-making. However, the unimodal nature of most current AI systems limits their ability to achieve a holistic understanding of surgical workflows. This highlights the need for general-purpose surgical AI systems capable of comprehensively modeling the interrelated components of surgical scenes. Recent advances in large vision-language models that integrate multimodal data processing offer strong potential for modeling surgical tasks and providing human-like scene reasoning and understanding. Despite their promise, systematic investigations of VLMs in surgical applications remain limited. In this study, we evaluate the effectiveness of large VLMs for the fundamental surgical vision task of detecting surgical tools. Specifically, we investigate three state-of-the-art VLMs, Qwen2.5, LLaVA1.5, and InternVL3.5, on the GraSP robotic surgery dataset under both zero-shot and parameter-efficient LoRA fine-tuning settings. Our results demonstrate that Qwen2.5 consistently achieves superior detection performance in both configurations among the evaluated VLMs. Furthermore, compared with the open-set detection baseline Grounding DINO, Qwen2.5 exhibits stronger zero-shot generalization and comparable fine-tuned performance. Notably, Qwen2.5 shows superior instrument recognition, while Grounding DINO demonstrates stronger localization.

arXiv:2501.02725v5 Announce Type: replace Abstract: Artificial intelligence (AI) has undergone transformative advances since 2022, particularly through generative AI, large language models (LLMs), and diffusion models, fundamentally reshaping the creative industries. However, existing reviews have not comprehensively addressed these recent breakthroughs and their integrated impact across the creative production pipeline. This paper addresses this gap by providing a systematic review of AI technologies that have emerged or matured since our 2022 review, examining their applications across content creation, information analysis, post-production enhancement, compression, and quality assessment. We document how transformers, LLMs, diffusion models, and implicit neural representations have established new capabilities in text-to-image/video generation, real-time 3D reconstruction, and unified multi-task frameworks-shifting AI from support tool to core creative technology. Beyond technological advances, we analyze the trend toward unified AI frameworks that integrate multiple creative tasks, replacing task-specific solutions. We critically examine the evolving role of human-AI collaboration, where human oversight remains essential for creative direction and mitigating AI hallucinations. Finally, we identify emerging challenges including copyright concerns, bias mitigation, computational demands, and the need for robust regulatory frameworks. This review provides researchers and practitioners with a comprehensive understanding of current AI capabilities, limitations, and future trajectories in creative applications.

arXiv:2506.11003v3 Announce Type: replace-cross Abstract: Large Language Models (LLMs) have shown promise in various tasks, yet few benchmarks assess their capabilities in embedded system development. In this paper, we introduce EmbedAgent, a paradigm designed to simulate real-world roles in embedded system development, such as Embedded System Programmer, Architect, and Integrator. This paradigm enables LLMs to be tested in tasks that bridge the gap between digital and physical systems, allowing for a more comprehensive assessment of their capabilities. To evaluate LLMs on these tasks, we propose Embedbench, the first comprehensive benchmark for embedded system programming, circuit design, and cross-platform migration. Embedbench consists of 126 cases, covering 9 electronic components across 3 hardware platforms. Through extensive experiments on 10 mainstream LLMs, we uncover several key findings. Surprisingly, despite the simplicity of the cases, DeepSeek-R1 achieves only a 55.6% pass@1 rate when provided with schematic information, and 50.0% when tasked with generating the schematics itself. In the cross-platform migration tasks, LLMs show relatively strong performance with MicroPython on the Raspberry Pi Pico (with the top model achieving 73.8% pass@1), but perform poorly on ESP-IDF, where the best model reaches only 29.4% pass@1. Interestingly, we observe that general-purpose chat LLMs like DeepSeek-V3 often fail to utilize relevant pre-trained knowledge in this domain, while reasoning LLMs tend to overthink and overlook efficient knowledge during pretraining. Based on these insights, we propose two strategies: retrieval augmented generation and compiler feedback-to enhance LLM performance. These strategies result in significant improvements, with Deepseek-R1 reaching a 65.1% pass@1 with correct schematics, and 53.1% without. Additionally, the accuracy of the Arduino to ESP32 migration task improves from 21.4% to 27.8%.

arXiv:2510.23845v2 Announce Type: replace-cross Abstract: Detecting mental health crisis situations such as suicide ideation, rape, domestic violence, child abuse, and sexual harassment is a critical yet underexplored challenge for language models. When such situations arise during user--model interactions, models must reliably flag them, as failure to do so can have serious consequences. In this work, we introduce CRADLE BENCH, a benchmark for multi-faceted crisis detection. Unlike previous efforts that focus on a limited set of crisis types, our benchmark covers seven types defined in line with clinical standards and is the first to incorporate temporal labels. Our benchmark provides 600 clinician-annotated evaluation examples and 420 development examples, together with a training corpus of around 4K examples automatically labeled using a majority-vote ensemble of multiple language models, which significantly outperforms single-model annotation. We further fine-tune six crisis detection models on subsets defined by consensus and unanimous ensemble agreement, providing complementary models trained under different agreement criteria.

npj Digital Medicine, Published online: 24 January 2026; doi:10.1038/s41746-026-02359-1

Deep learning for malignancy and tumor origin prediction using cytology or histopathology whole slide images

npj Digital Medicine, Published online: 24 January 2026; doi:10.1038/s41746-025-02320-8

The diagnostic accuracy of wearable digital technology in detecting fertility window and menstrual cycles: a systematic review and Bayesian network meta-analysis

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.

“Dr. Google” had its issues. Can ChatGPT Health do better?  

For the past two decades, there’s been a clear first step for anyone who starts experiencing new medical symptoms: Look them up online. The practice was so common that it gained the pejorative moniker “Dr. Google.” But times are changing, and many medical-information seekers are now using LLMs. According to OpenAI, 230 million people ask ChatGPT health-related queries each week.  

That’s the context around the launch of OpenAI’s new ChatGPT Health product, which debuted earlier this month. The big question is: can the obvious risks of using AI for health-related queries be mitigated enough for them to be a net benefit? Read the full story. 

—Grace Huckins

America’s coming war over AI regulation  

In the final weeks of 2025, the battle over regulating artificial intelligence in the US reached boiling point. On December 11, after Congress failed twice to pass a law banning state AI laws, President Donald Trump signed a sweeping executive order seeking to handcuff states from regulating the booming industry.  

Instead, he vowed to work with Congress to establish a “minimally burdensome” national AI policy. The move marked a victory for tech titans, who have been marshaling multimillion-dollar war chests to oppose AI regulations, arguing that a patchwork of state laws would stifle innovation.

In 2026, the battleground will shift to the courts. While some states might back down from passing AI laws, others will charge ahead. Read our story about what’s on the horizon. 

—Michelle Kim

This story is from MIT Technology Review’s What’s Next series of stories that look across industries, trends, and technologies to give you a first look at the future. You can read the rest of them here.  

Measles is surging in the US. Wastewater tracking could help.

This week marked a rather unpleasant anniversary: It’s a year since Texas reported a case of measles—the start of a significant outbreak that ended up spreading across multiple states. Since the start of January 2025, there have been over 2,500 confirmed cases of measles in the US. Three people have died. 

As vaccination rates drop and outbreaks continue, scientists have been experimenting with new ways to quickly identify new cases and prevent the disease from spreading. And they are starting to see some success with wastewater surveillance. Read the full story.

—Jessica Hamzelou 

This story is from The Checkup, our weekly newsletter giving you the inside track on all things health and biotech. Sign up to receive it in your inbox every Thursday.

The must-reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 The US is dismantling itself
A foreign enemy could not invent a better chain of events to wreck its standing in the world. (Wired $)  
+ We need to talk about whether Donald Trump might be losing it.  (New Yorker $)

2 Big Tech is taking on more debt to fund its AI aspirations
And the bubble just keeps growing. (WP $)
+ Forget unicorns. 2026 is shaping up to be the year of the “hectocorn.” (The Guardian)
+ Everyone in tech agrees we’re in a bubble. They just can’t agree on what happens when it pops. (MIT Technology Review)

3 DOGE accessed even more personal data than we thought 
Even now, the Trump administration still can’t say how much data is at risk, or what it was used for. (NPR)

4 TikTok has finalized a deal to create a new US entity 
Ending years of uncertainty about its fate in America. (CNN)
+ Why China is the big winner out of all of this. (FT $)

5 The US is now officially out of the World Health Organization 
And it’s leaving behind nearly $300 million in bills unpaid. (Ars Technica) 
+ The US withdrawal from the WHO will hurt us all. (MIT Technology Review)

6 AI-powered disinformation swarms pose a threat to democracy
A would-be autocrat could use them to persuade populations to accept cancelled elections or overturn results. (The Guardian)
+ The era of AI persuasion in elections is about to begin. (MIT Technology Review)

7 We’re about to start seeing more robots everywhere
But exactly what they’ll look like remains up for debate. (Vox $)
+ Chinese companies are starting to dominate entire sectors of AI and robotics. (MIT Technology Review)

8 Some people seem to be especially vulnerable to loneliness
If you’re ‘other-directed’, you could particularly benefit from less screentime. (New Scientist $)

9 This academic lost two years of work with a single click
TL;DR: Don’t rely on ChatGPT to store your data. (Nature)

10 How animals develop a sense of direction 
Their ‘internal compass’ seems to be informed by landmarks that help them form a mental map. (Quanta $)

Quote of the day

“The rate at which AI is progressing, I think we have AI that is smarter than any human this year, and no later than next year.”

—Elon Musk simply cannot resist the urge to make wild predictions at Davos, Wired reports. 

One more thing

Africa fights rising hunger by looking to foods of the past

After falling steadily for decades, the prevalence of global hunger is now on the rise—nowhere more so than in sub-Saharan Africa. 

Africa’s indigenous crops are often more nutritious and better suited to the hot and dry conditions that are becoming more prevalent, yet many have been neglected by science, which means they tend to be more vulnerable to diseases and pests and yield well below their theoretical potential.

Now the question is whether researchers, governments, and farmers can work together in a way that gets these crops onto plates and provides Africans from all walks of life with the energy and nutrition that they need to thrive, whatever climate change throws their way. Read the full story.

—Jonathan W. Rosen

We can still have nice things

A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.)

+ The only thing I fancy dry this January is a martini. Here’s how to make one.
+ If you absolutely adore the Bic crystal pen, you might want this lamp. 
+ Cozy up with a nice long book this winter. ($)
+ Want to eat healthier? Slow down and tune out food ‘noise’. ($)

Nature Medicine, Published online: 23 January 2026; doi:10.1038/s41591-025-04198-1

We propose straightforward principles to foster an evaluation-forward operating system that can transform the adoption of clinical artificial intelligence from a leap of faith into a stepwise, trust-building process.