Transduction efficiency is a critical determinant of CAR T-cell manufacturing quality. Analysis of 204 clinical CAR T-cell products revealed that transduction efficiency is shaped primarily by manufacturing workflows and protocol-dependent starting material composition. Higher transduction efficiency was associated with early memory-like cellular states, providing insights into optimizing CAR T-cell.

He and colleagues developed a transient, non-gene-editing platform for off-the-shelf allogeneic TCR-T therapy. By conferring tacrolimus resistance characteristics to IL-2/4/7 expanded TCR-T cells that display reduced alloreactivity, He et al. provided a proof-of-concept for an allogeneic TCR-T therapy approach that addresses both manufacturability, and unwanted graft and host alloreactivity concerns.

Small extracellular vesicles from differentiating human skeletal muscle cells are enriched in miR-193b-3p, which contributes to myogenic responses through AP2M1 regulation. These findings establish a miR-193b-3p–AP2M1 axis connecting extracellular vesicles to muscle regeneration and highlight its potential role in regulating myogenic processes.

Et1Ψ-modified mRNA shows UUU-dependent translational sensitivity, but targeted UUU-to-UUC recoding restores protein expression. With this sequence constraint addressed, Et1Ψ reduces early innate immune activation while preserving vaccine-induced humoral, cellular, and neutralizing responses, thereby establishing codon–modification compatibility as a practical design principle for mRNA therapeutics.

Oncolytic herpes simplex virus 1 (oHSV) is a promising viro-immunotherapy that directly lyses tumor cells while reshaping the tumor microenvironment (TME) and stimulates anti-tumor immunity. However, despite encouraging preclinical and clinical activity, treatment-induced adaptations in tumor cells and the surrounding microenvironment can promote survival, repair, repopulation, immune escape, and recurrence, thereby limiting durable therapeutic benefit. In our recent article published in Cell Death & Disease,1 we asked why oHSV often fails to achieve durable tumor control and whether the adaptive resistance it induces can be therapeutically overcome.

Yuzaki and colleagues identify biallelic CBLN1 variants as a cause of early-onset hereditary ataxia and show that loss of extracellular CBLN1 disrupts cerebellar synapses. Astrocyte-targeted AAV delivery restores synaptic CBLN1 and rescues circuit and motor dysfunction, establishing extracellular synaptic organizer replacement as a therapeutic strategy.

Lee and colleagues identify MFT251 as a synthetic small-molecule TBK1 agonist that directly activates type I interferon signaling. MFT251 induces broad antiviral defenses and reshapes the tumor microenvironment to enhance antitumor immunity, establishing pharmacologic TBK1 activation as a strategy for innate immune modulation in infection and cancer.

Red blood cell-derived extracellular vesicles (RBCEVs) provide a scalable, non-viral platform for gene therapy. High-grade RBCEVs purified using tangential flow filtration can deliver sizable plasmids and mediate sustained in vivo expression of therapeutic proteins, including factor IX and Herceptin, highlighting their potential use for safe and efficient gene therapy.

Liu and colleagues developed a robust in cellulo plasmid DNA replication system in human cells for replicating plasmid-borne adeno-associated virus (AAV) Rep/Cap genes during recombinant AAV (rAAV) production. This new approach not only enables a 10- to 20-fold plasmid reduction to significantly lower manufacturing costs but also substantially enhances rAAV potency, titer, and purity.

Song and colleagues report an enterovirus-derived capless self-amplifying mRNA vaccine platform that uses internal ribosome entry site-mediated translation to eliminate 5′ capping. When encoding H5 hemagglutinin, CLsamRNA induces robust humoral and cellular immunity and dose-sparing protection against lethal H5N8 challenge in mice, supporting a distinct RNA vaccine architecture.

Lee and colleagues develop anti-TLR2 nanobodies and scFvs and establish a PLGA-nanoparticle-mediated gene therapy platform for sustained antibody expression. A single intrathecal injection of these PLGA nanoparticles reduces spinal inflammatory responses and provides durable analgesia in a mouse model of neuropathic pain.

Subretinal fibrosis causes irreversible vision loss in neovascular age-related macular degeneration (AMD). This study shows that proline metabolism, particularly P4HA1-mediated proline hydroxylation, is activated in JR5558 mice and human AMD tissues. Diethyl pythiDC reduced collagen turnover and fibrovascular lesion expansion, with potential added benefit when combined with aflibercept.

CAR T cells engineered to secrete tumor-localized anti-EpCAM bispecific T cell engagers (BTCEs) overcome antigen escape and heterogeneity across multiple epithelial carcinomas in preclinical models, achieving complete tumor eradication where conventional single-target CAR T cell therapies often failed, supporting broad translational potential for solid tumor immunotherapy.

This article demonstrates the use of adenine base editor nickase activity for precise transgene cargo integration while simultaneously carrying out multiplex gene knockout. This single-step platform achieves efficient knockin and >90% quadplex gene knockout without detectable translocations and low levels of off-target editing, producing highly functional, customizable chimeric antigen receptor T cells.

Nature Nanotechnology, Published online: 03 September 2026; doi:10.1038/s41565-026-02267-1

Impurities in a quantum exciton condensate create bound electronic states, as visualized in Ta2Pd3Te5, that are analogous to spin impurities in superconductors, enabling microscopic control of excitonic quantum matter.

arXiv:2609.09324v1 Announce Type: cross Abstract: Link Adaptation (LA) in 5G NR is inherently reactive, relying on channel measurements and HARQ feedback that may become quickly obsolete when the channel changes quickly. This data is also noisy, making it hard to track accurately, and has to be fed to real-time controllers with feedback-loop effects which are hard to troubleshoot. This explains why most practical deployments select simple but robust algorithms, which accept that the lag can leave the scheduler operating at overly aggressive or unnecessarily conservative rates, trading spectrum efficiency for predictable performance. In this paper, we improve on this status-quo with NOSTRAdAMUS, a predictive LA framework which adds foresight to existing algorithms without replacing or redesigning them. NOSTRAdAMUS predicts whether a retransmission will occur in the next radio frame from recent HARQ history, and applies corrections to the Modulation and Coding Scheme (MCS) selected by the underlying policy. We benchmark several ML models and show that Gradient Boosting achieves 82.9% accuracy overall with high-confidence interventions that are correct 94.2% of the time, and an inference latency of 5.5 {\mu}s. We train the model based on data collected Over-the-Air (OTA) on the X5G testbed, using the open-source OpenAirInterface (OAI) 5G stack, NVIDIA Aerial, and COTS O-RAN Radio Units and User Equipments. The model is then deployed as a dApp, which we evaluate OTA as well as on various channels with hardware-in-the-loop channel emulators. This includes 3GPP TDL and CDL channels, SISO and MIMO configurations, and pedestrian and vehicular mobility. Our evaluation shows that without retraining, and across this variety of scenarios, the dApp augments two SOTA LA algorithms, and increases goodput by up to 71.5% while reducing retransmissions by up to 71.8%. This demonstrates the robustness and generalization capabilities of our approach.

arXiv:2609.10052v1 Announce Type: cross Abstract: LLM agents for sequential decision tasks are often post-trained with trajectory-level outcome labels, but such labels provide little supervision for preserving multiple successful branches from the same decision state. We study this problem as successful strategy coverage: how broadly a model realizes distinct successful strategies under a fixed rollout budget. We present Direct Diversity Optimization (DDO), an offline post-training method that combines Divergence-Tree Collection (DTC) with the Reference-Relative Target-Odds Objective (RTO). DTC constructs state-aligned branch sets rooted at shared decision states, and RTO trains the model to match reference-relative targets over successful alternatives. DDO achieves the strongest task success and successful strategy coverage among the compared post-training methods across BabyAI, BabaIsAI, and WebShop. It also achieves the highest recovery rate after local action replacement and higher task success and coverage than successful-only imitation and decoding-time diversification controls.

arXiv:2609.10410v1 Announce Type: cross Abstract: The growing complexity of content moderation policies presents a critical challenge for their consistent operationalization. While foundation models possess the basic capabilities needed to confront this challenge, whether they can reliably moderate online content remains an unanswered question. In this paper, we systematically compare two competing paradigms for Vision-Language Model (VLM) guidance: an instruction-driven approach where models reason from policy precepts, and an example-driven approach where they generalize from prior precedents. We ground this investigation in ModerationBench, a new benchmark of 4,000 manually annotated, in-the-wild posts from the Bluesky platform. Our experiments reveal that foundation models can substantially outperform Bluesky's deployed moderation system, nearly tripling its $F_1$ score (0.60 vs. 0.22) on Random Posts in the benchmark, with both instruction- and example-driven paradigms achieving comparable peak effectiveness. Our findings thus chart a path toward reliable and adaptable policy operationalization at scale.

arXiv:2609.10421v1 Announce Type: cross Abstract: Background: Emergency Department (ED) return visits are commonly reviewed for quality assurance, but are often limited (e.g., to revisits within 48-72 hours) to increase actionable finding yield while minimizing chart review burden. Those limitations may lead to missed quality improvement opportunities. Methods: We conducted an exploratory, retrospective study of randomly selected ED visits to a multihospital health system having an ED revisit within 1-14 days to the same health system. Given only each visit's primary diagnosis, raters (2-3 clinicians and GPT-4 large language model [LLM]) assessed characteristics of the diagnosis pairs, including the "target": whether a pair warranted further assessment. Informed by rater response analyses, an algorithm leveraging an LLM-populated knowledge graph ("KGA") was created to automatically screen for potentially concerning pairs, then preliminarily assessed. Results: 99 diagnosis pairs were included. GPT-4 responses poorly correlated to clinician raters, rating nearly all (94%) pairs as warranting follow-up (4.4-13.3 times more than clinicians). However, prompt engineering was minimal. Among clinician raters, revisit medical gravity was consistently significantly associated with the target, while a differential diagnosis/complication composite was significantly associated on unadjusted, but not adjusted (though less powered) analysis. The KGA achieved 83-100% positive predictive value for at least one clinician rater determining further assessment was warranted based on the diagnosis pair. Conclusion: These results can inform next steps for improving screening with LLMs like ChatGPT. Further research is warranted to validate this preliminary work's finding that the KGA may enable enhancing the scope and yield of screening without substantially increasing reviewer workload.

arXiv:2604.10333v2 Announce Type: replace Abstract: Young children demonstrate early abilities to understand their physical world, estimating depth, motion, object coherence, interactions, and many other aspects of physical scene understanding. Children are both data-efficient and flexible cognitive systems, creating competence despite extremely limited training data, while generalizing to myriad untrained tasks -- a major challenge even for today's best AI systems. Here we introduce a novel computational hypothesis for these abilities, the Zero-shot World Model (ZWM). ZWM is based on three principles: a sparse temporally-factored predictor that decouples appearance from dynamics; zero-shot estimation through approximate causal inference; and composition of inferences to build more complex abilities. We show that ZWM can be learned from the first-person experience of a single child, rapidly generating competence across multiple physical understanding benchmarks. It also shows progressive, staged emergence of capacities during learning and builds brain-like internal representations. Our work presents a blueprint for efficient and flexible learning from human-scale data, advancing both a computational account of children's early physical understanding and a path toward data-efficient AI systems.