arXiv:2510.19033v1 Announce Type: cross Abstract: While much research has shown the presence of AI's "under-the-hood" biases (e.g., algorithmic, training data, etc.), what about "over-the-hood" inclusivity biases: barriers in user-facing AI products that disproportionately exclude users with certain problem-solving approaches? Recent research has begun to report the existence of such biases -- but what do they look like, how prevalent are they, and how can developers find and fix them? To find out, we conducted a field study with 3 AI product teams, to investigate what kinds of AI inclusivity bugs exist uniquely in user-facing AI products, and whether/how AI product teams might harness an existing (non-AI-oriented) inclusive design method to find and fix them. The teams' work resulted in identifying 6 types of AI inclusivity bugs arising 83 times, fixes covering 47 of these bug instances, and a new variation of the GenderMag inclusive design method, GenderMag-for-AI, that is especially effective at detecting certain kinds of AI inclusivity bugs.

arXiv:2510.19181v1 Announce Type: cross Abstract: This paper presents a question answering system that operates exclusively on a knowledge graph retrieval without relying on retrieval augmented generation (RAG) with large language models (LLMs). Instead, a small paraphraser model is used to paraphrase the entity relationship edges retrieved from querying the knowledge graph. The proposed pipeline is divided into two main stages. The first stage involves pre-processing a document to generate sets of question-answer (QA) pairs. The second stage converts these QAs into a knowledge graph from which graph-based retrieval is performed using embeddings and fuzzy techniques. The graph is queried, re-ranked, and paraphrased to generate a final answer. This work includes an evaluation using LLM-as-a-judge on the CRAG benchmark, which resulted in accuracies of 71.9% and 54.4% using LLAMA-3.2 and GPT-3.5-Turbo, respectively.

arXiv:2510.19535v1 Announce Type: cross Abstract: AI methods are increasingly shaping pharmaceutical drug discovery. However, their translation to industrial applications remains limited due to their reliance on public datasets, lacking scale and diversity of proprietary pharmaceutical data. Federated learning (FL) offers a promising approach to integrate private data into privacy-preserving, collaborative model training across data silos. This federated data access complicates important data-centric tasks such as estimating dataset diversity, performing informed data splits, and understanding the structure of the combined chemical space. To address this gap, we investigate how well federated clustering methods can disentangle and represent distributed molecular data. We benchmark three approaches, Federated kMeans (Fed-kMeans), Federated Principal Component Analysis combined with Fed-kMeans (Fed-PCA+Fed-kMeans), and Federated Locality-Sensitive Hashing (Fed-LSH), against their centralized counterparts on eight diverse molecular datasets. Our evaluation utilizes both, standard mathematical and a chemistry-informed evaluation metrics, SF-ICF, that we introduce in this work. The large-scale benchmarking combined with an in-depth explainability analysis shows the importance of incorporating domain knowledge through chemistry-informed metrics, and on-client explainability analyses for federated diversity analysis on molecular data.

arXiv:2510.19799v1 Announce Type: cross Abstract: Public and nonprofit organizations often hesitate to adopt AI tools because most models are opaque even though standard approaches typically analyze aggregate patterns rather than offering actionable, case-level guidance. This study tests a practitioner-in-the-loop workflow that pairs transparent decision-tree models with large language models (LLMs) to improve predictive accuracy, interpretability, and the generation of practical insights. Using data from an ongoing college-success program, we build interpretable decision trees to surface key predictors. We then provide each tree's structure to an LLM, enabling it to reproduce case-level predictions grounded in the transparent models. Practitioners participate throughout feature engineering, model design, explanation review, and usability assessment, ensuring that field expertise informs the analysis at every stage. Results show that integrating transparent models, LLMs, and practitioner input yields accurate, trustworthy, and actionable case-level evaluations, offering a viable pathway for responsible AI adoption in the public and nonprofit sectors.

arXiv:2510.16206v2 Announce Type: replace Abstract: Since the rapid expansion of large language models (LLMs), people have begun to rely on them for information retrieval. While traditional search engines display ranked lists of sources shaped by search engine optimization (SEO), advertising, and personalization, LLMs typically provide a synthesized response that feels singular and authoritative. While both approaches carry risks of bias and omission, LLMs may amplify the effect by collapsing multiple perspectives into one answer, reducing users ability or inclination to compare alternatives. This concentrates power over information in a few LLM vendors whose systems effectively shape what is remembered and what is overlooked. As a result, certain narratives, individuals or groups, may be disproportionately suppressed, while others are disproportionately elevated. Over time, this creates a new threat: the gradual erasure of those with limited digital presence, and the amplification of those already prominent, reshaping collective memory. To address these concerns, this paper presents a concept of the Right To Be Remembered (RTBR) which encompasses minimizing the risk of AI-driven information omission, embracing the right of fair treatment, while ensuring that the generated content would be maximally truthful.

arXiv:2503.12730v5 Announce Type: replace-cross Abstract: Mechanistic interpretability research faces a gap between analyzing simple circuits in toy tasks and discovering features in large models. To bridge this gap, we propose text-to-SQL generation as an ideal task to study, as it combines the formal structure of toy tasks with real-world complexity. We introduce TinySQL, a synthetic dataset, progressing from basic to advanced SQL operations, and train models ranging from 33M to 1B parameters to establish a comprehensive testbed for interpretability. We apply multiple complementary interpretability techniques, including Edge Attribution Patching and Sparse Autoencoders, to identify minimal circuits and components supporting SQL generation. We compare circuits for different SQL subskills, evaluating their minimality, reliability, and identifiability. Finally, we conduct a layerwise logit lens analysis to reveal how models compose SQL queries across layers: from intent recognition to schema resolution to structured generation. Our work provides a robust framework for probing and comparing interpretability methods in a structured, progressively complex setting.

arXiv:2504.09909v2 Announce Type: replace-cross Abstract: In recent developments, deep learning methodologies applied to Natural Language Processing (NLP) have revealed a paradox: They improve performance but demand considerable data and resources for their training. Alternatively, quantum computing exploits the principles of quantum mechanics to overcome the computational limitations of current methodologies, thereby establishing an emerging field known as quantum natural language processing (QNLP). This domain holds the potential to attain a quantum advantage in the processing of linguistic structures, surpassing classical models in both efficiency and accuracy. In this paper, it is proposed to categorise QNLP models based on quantum computing principles, architecture, and computational approaches. This paper attempts to provide a survey on how quantum meets language by mapping state-of-the-art in this area, embracing quantum encoding techniques for classical data, QNLP models for prevalent NLP tasks, and quantum optimisation techniques for hyper parameter tuning. The landscape of quantum computing approaches applied to various NLP tasks is summarised by showcasing the specific QNLP methods used, and the popularity of these methods is indicated by their count. From the findings, it is observed that QNLP approaches are still limited to small data sets, with only a few models explored extensively, and there is increasing interest in the application of quantum computing to natural language processing tasks.

arXiv:2505.07897v3 Announce Type: replace-cross Abstract: Context lengths for models have grown rapidly, from thousands to millions of tokens in just a few years. The extreme context sizes of modern long-context models have made it difficult to construct realistic long-context benchmarks -- not only due to the cost of collecting million-context tasks but also in identifying realistic scenarios that require significant contexts. We identify code comprehension and repair as a natural testbed and challenge task for long-context models and introduce LongCodeBench (LCB), a benchmark to test LLM coding abilities in long-context scenarios. Our benchmark tests both the comprehension and repair capabilities of LCLMs in realistic and important settings by drawing from real-world GitHub issues and constructing QA (LongCodeQA) and bug fixing (LongSWE-Bench) tasks. We carefully stratify the complexity of our benchmark, enabling us to evaluate models across different scales -- ranging from Qwen2.5 14B Instruct to Google's flagship Gemini model. We find that long-context remains a weakness for all models, with performance drops such as from 29% to 3% for Claude 3.5 Sonnet, or from 70.2% to 40% for Qwen2.5. The LCB dataset is available publicly at https://huggingface.co/datasets/Steefano/LCB and the codebase to replicate the work on this paper at https://github.com/Zteefano/long-code-bench.

arXiv:2506.16895v2 Announce Type: replace-cross Abstract: Multimodal models have demonstrated powerful capabilities in complex tasks requiring multimodal alignment, including zero-shot classification and cross-modal retrieval. However, existing models typically rely on millions of paired multimodal samples, which are prohibitively expensive or infeasible to obtain in many domains. In this work, we explore the feasibility of building multimodal models with limited amount of paired data by aligning pretrained unimodal foundation models. We show that high-quality alignment is possible with as few as tens of thousands of paired samples$\unicode{x2013}$less than $1\%$ of the data typically used in the field. To achieve this, we introduce STRUCTURE, an effective regularization technique that preserves the neighborhood geometry of the latent space of unimodal encoders. Additionally, we show that aligning last layers is often suboptimal and demonstrate the benefits of aligning the layers with the highest representational similarity across modalities. These two components can be readily incorporated into existing alignment methods, yielding substantial gains across 24 zero-shot image classification and retrieval benchmarks, with average relative improvement of $51.6\%$ in classification and $91.8\%$ in retrieval tasks. Our results highlight the effectiveness and broad applicability of our framework for limited-sample multimodal learning and offer a promising path forward for resource-constrained domains.

arXiv:2509.26433v2 Announce Type: replace-cross Abstract: When used in high-stakes settings, AI systems are expected to produce decisions that are transparent, interpretable, and auditable, a requirement increasingly expected by regulations. Decision trees such as CART provide clear and verifiable rules, but they are restricted to structured tabular data and cannot operate directly on unstructured inputs such as text. In practice, large language models (LLMs) are widely used for such data, yet prompting strategies such as chain-of-thought or prompt optimization still rely on free-form reasoning, limiting their ability to ensure trustworthy behaviors. We present the Agentic Classification Tree (ACT), which extends decision-tree methodology to unstructured inputs by formulating each split as a natural-language question, refined through impurity-based evaluation and LLM feedback via TextGrad. Experiments on text benchmarks show that ACT matches or surpasses prompting-based baselines while producing transparent and interpretable decision paths.

Moderna said Wednesday afternoon that its experimental vaccine for cytomegalovirus, a cause of disability in newborns, failed in a Phase 3 trial, a significant setback for a company already facing pressure from Wall Street and the federal government.

The CMV vaccine had been the company’s lead program prior to the Covid-19 pandemic. Leadership had repeatedly said it could bring in between $2 billion and $5 billion in peak annual sales. Analysts polled by Visible Alpha forecast peak sales of $1.6 billion for the product.

“It’s obviously disappointing,” said Stephen Hoge, Moderna’s president, in an interview.

Continue to STAT+ to read the full story…

© Ruby Wallau for STAT

BERLIN — The leading professional organization for European oncologists has rolled out its first set of guidance on how its members should use large language models, a type of artificial intelligence, in cancer medicine. 

“The oncology community cannot ignore the potential benefits which AI technology can provide to cancer patients,” the authors of the guidance wrote, while simultaneously acknowledging that there aren’t enough evaluations of the chatbots available to patients or tools available to doctors to address the risks associated with generative AI in medicine.

The guidance’s release — it was published last Saturday in the Annals of Oncology — coincided with the annual meeting for the European Society for Clinical Oncology in Berlin. The American Society of Clinical Oncology has issued its own set of principles for the responsible use of AI in cancer medicine, but has not released recommendations specific to large language models (LLMs).

Continue to STAT+ to read the full story…

© Adobe

Nature, Published online: 22 October 2025; doi:10.1038/s41586-025-09761-x

Discovering state-of-the-art reinforcement learning algorithms

Nature, Published online: 22 October 2025; doi:10.1038/d41586-025-03133-1

Communication between epithelial and mesenchymal cells in pancreatic cancer leads to a poor prognosis. The molecular basis for this signalling has now been revealed.

Oncogene, Published online: 21 October 2025; doi:10.1038/s41388-025-03605-8

Meflin is a druggable target using antibody-drug conjugates in progressive osteosarcoma

npj Digital Medicine, Published online: 21 October 2025; doi:10.1038/s41746-025-02082-3

Although still at a nascent state, quantum computing promises advances in healthcare, from drug discovery to personalised treatments. But it also threatens current cryptographic systems that protect medical data and infrastructure. The concept of “Q-Day” highlights risks such as “harvest now, decrypt later” attacks, with particular concerns for medical devices and sensitive applications in fields like femtech. Preparing for this future requires the rapid adoption of post-quantum cryptography, the coordination of time-phased and scalable “technology rollout” strategies, and revised regulatory frameworks to safeguard patient safety, privacy, and trust.

Nature Biotechnology, Published online: 21 October 2025; doi:10.1038/s41587-025-02856-w

Multiplexed editing in primary human T cells generates enhanced immune cell therapies.

You’re reading the web edition of STAT’s Health Tech newsletter, our guide to how technology is transforming the life sciences. Sign up to get it delivered in your inbox every Tuesday and Thursday.

Abridge vs. OpenEvidence, round one

New announcements from Abridge and OpenEvidence,  two of the most prominent health tech startups to emerge in recent years, highlight how despite different initial offerings, many artificial intelligence companies in health care will just end up competing with each other for physician eyeballs.

On Monday morning, Abridge, best known for its AI scribe that helps doctors automate the writing of clinical notes, announced a new product that will surface “real-time insights, prompts, and pathways” from the widely-used medical resource UpToDate based on things said in a clinical conversation and that are written in the patient’s record. 

Continue to STAT+ to read the full story…

© ADOBE, Alex Hogan/STAT

Background: Obesity is a chronic, relapsing disease influenced by environmental, lifestyle, biological, and genetic factors, affecting over 1 billion people globally. Treatment for adults typically involves multicomponent lifestyle interventions—diet, physical activity, and behavior change—for at least 6-12 months. However, adherence is often low, and in-person sessions can be time-consuming and costly. Digital therapeutics (DTx), which enhance patient engagement and support long-term outcomes, have proven effective in managing chronic and mental health conditions. DTx offer scalable, evidence-based solutions with the potential to improve obesity management. Objective: The Digital Therapy to Promote Weight Loss in Patients With Obesity by Increasing Their Adherence to Treatment (DEMETRA) study is a prospective, multicenter, pragmatic, randomized, double-arm, single-blind, placebo-controlled trial evaluating the 6-month efficacy of an innovative, multicomponent digital intervention for obesity, which combines dietary, physical activity, and behavioral strategies in people with obesity (primary objective). Secondary objectives were assessing changes in BMI, waist circumference, blood pressure, glucose metabolism, lipid profile, adherence, and factors associated with absolute 6-month weight loss. Methods: The trial was conducted at 2 obesity centers in Italy with 246 participants aged 18-65 years (BMI 30-45 kg/m2), randomly assigned to either the Digital Therapeutics for Obesity (DTxO) app or a placebo app. DTxO offered personalized diet plans, exercise routines, and psycho-behavioral support, while the placebo app only allowed users to log data without feedback. Both groups followed a Mediterranean-style low-calorie diet with an 800 kcal/day deficit. On average, participants used the DTxO app for 42 minutes/day and the placebo app for 35 minutes, primarily for physical activity tracking. Univariable and multivariable generalized linear models were used to assess associations with 6-month absolute weight change (primary end point) and percent weight change (secondary end point). Results: Overall, 207 participants (84.1%) completed the 6-month visit. Both arms achieved a statistically significant absolute (DtxO: –3.2 kg, IQR –6.0 kg to –0.9 kg; placebo: –4.0 kg, IQR –6.9 kg to –0.5 kg; P<.001) and percent loss in body weight (DtxO: –3.0%, IQR –5.7% to –0.8%; placebo: –4.0%, IQR –8.5% to –0.5%; P<.001) after 6 months, without significant between-group differences (univariable generalized linear models: P=.34 and P=.17, respectively). Univariable regression analyses showed a significant association between adherence to app use and 6-month absolute weight loss (β=–.06, SE 0.02, P=.01) as well as percent weight loss (β=–.05, SE 0.01, P=.01). Adherent participants, defined as those with overall adherence at or above the 75th percentile of daily usage, included 35 individuals in the intervention group and 10 in the placebo group. In this subgroup, the estimated 6-month mean absolute weight change was –7.02 kg (95% CI –9.45 to –4.59) in the DTxO-adherent group and –3.50 kg (95% CI –7.01 to 0.01) in the placebo-adherent group (P=.02). The estimated 6-month mean percent change in weight was –6.31% (95% CI –8.86 to –3.76) in the DTxO-adherent group and –2.78% (95% CI –6.48 to 0.92) in the placebo-adherent group (P=.03). A significantly greater weight loss (P=.01 for study arm, either on absolute or percent change in weight from baseline) among adherent participants randomized to the DTxO app was also confirmed by analyses using mixed linear models for repeated measures. Conclusions: Although overall weight loss did not differ significantly between the DTxO and placebo groups, participants who used the DTxO app for at least 40% of the expected time achieved significantly greater weight loss. These results suggest that higher engagement with DTx can improve obesity outcomes. Further research should explore combining DTxO with pharmacological treatments or bariatric surgery. Trial Registration: ClinicalTrials.gov NCT05394779; https://clinicaltrials.gov/ct2/show/NCT05394779