Your personalresearch library

Isabelle Albuquerque Reis, Lilia Maria Lima de Oliveira, Fátima Maria Thaiz da Fonte Gomes da Silva, Lyria de Oliveira Rosa, Gabrielle de Lacerda Dantas Henrique, Nathália Freire Borba, Larissa Calixto Hespanhol, Heloísa Carneiro Brito, João Vitor Andrade Fernandes, Beatriz Ximenes Mendes, Naiara Abreu Fraga‐Braghiroli

JDDG: Journal der Deutschen Dermatologischen Gesellschaft·2025

Lip squamous cell carcinoma (SCC) accounts for approximately 30% of oral malignancies. Delayed diagnosis worsens the prognosis, leading to tumor growth, regional lymph node involvement, distant metastasis, and the need for invasive therapy. Dermoscopy is a practical, low‐cost, and non‐invasive diagnostic tool used to detect skin tumors. This review analyzed the common dermoscopic features of lip SCC and actinic cheilitis (AC). A search of MEDLINE, EMBASE, and Web of Science was made for original dermatoscopy studies. Pooled proportions were calculated using a random‐effects model (95% CI). Eleven studies were included: 325 patients, 205 lip SCCs, and 122 ACs. Key dermoscopic findings in lip SCCs were white structureless areas 61.9%, ulceration 59.3%, white halos 57.7%, keratin scales 54.98%, pinkish‐red background 50.5%, whitish‐yellow colors 41.8%, and shiny white streaks 24.5%. Polymorphous vessels were found in 55.8%, and monomorphous in 44.1%. Vessel morphology was hairpin 30.90%, serpentine 22.27%, dotted 21.46%, and glomerular 11.20%. ACs exhibited white halos in 66.14% of cases, white structureless areas in 59.62%, white scales in 57.90%, yellow scales in 30.14%, ulceration in 32.43%, white circles in 9.94%, and had polymorphous (72.93%), hairpin (30.83%), or dotted (20.71%) vessels. Dermoscopy aids early detection of lip SCC and AC, though overlapping features can limit accurate distinction. Histopathological confirmation remains essential in difficult cases, and further studies should explore subgroup‐specific dermoscopic patterns.

★ 5.0 (1)

View paper →

Page 1 of 374Next

Next-Embedding Prediction Makes Strong Vision Learners

Sihan Xu, Ziqiao Ma, Wenhao Chai, Xuweiyi Chen, Weiyang Jin, Joyce Chai, Saining Xie, Stella X. Yu

arXiv·2025

Inspired by the success of generative pretraining in natural language, we ask whether the same principles can yield strong self-supervised visual learners. Instead of training models to output features for downstream use, we train them to generate embeddings to perform predictive tasks directly. This work explores such a shift from learning representations to learning models. Specifically, models learn to predict future patch embeddings conditioned on past ones, using causal masking and stop gradient, which we refer to as Next-Embedding Predictive Autoregression (NEPA). We demonstrate that a simple Transformer pretrained on ImageNet-1k with next embedding prediction as its sole learning objective is effective - no pixel reconstruction, discrete tokens, contrastive loss, or task-specific heads. This formulation retains architectural simplicity and scalability, without requiring additional design complexity. NEPA achieves strong results across tasks, attaining 83.8% and 85.3% top-1 accuracy on ImageNet-1K with ViT-B and ViT-L backbones after fine-tuning, and transferring effectively to semantic segmentation on ADE20K. We believe generative pretraining from embeddings provides a simple, scalable, and potentially modality-agnostic alternative to visual self-supervised learning.

No ratings yet

View paper →

Distributional AGI Safety

Nenad Tomašev, Matija Franklin, Julian Jacobs, Sébastien Krier, Simon Osindero

arXiv·2025

AI safety and alignment research has predominantly been focused on methods for safeguarding individual AI systems, resting on the assumption of an eventual emergence of a monolithic Artificial General Intelligence (AGI). The alternative AGI emergence hypothesis, where general capability levels are first manifested through coordination in groups of sub-AGI individual agents with complementary skills and affordances, has received far less attention. Here we argue that this patchwork AGI hypothesis needs to be given serious consideration, and should inform the development of corresponding safeguards and mitigations. The rapid deployment of advanced AI agents with tool-use capabilities and the ability to communicate and coordinate makes this an urgent safety consideration. We therefore propose a framework for distributional AGI safety that moves beyond evaluating and aligning individual agents. This framework centers on the design and implementation of virtual agentic sandbox economies (impermeable or semi-permeable), where agent-to-agent transactions are governed by robust market mechanisms, coupled with appropriate auditability, reputation management, and oversight to mitigate collective risks.

No ratings yet

View paper →

Recent discoveries on the acquisition of the highest levels of human performance

Arne Güllich, Michael Barth, David Z. Hambrick, Brooke N. Macnamara

Science·2025

Scientists have long debated the origins of exceptional human achievements. This literature review summarizes recent evidence from multiple domains on the acquisition of world-class performance. We review published papers and synthesize developmental patterns of international top scientists, musicians, athletes, and chess players. The available evidence is highly consistent across domains: (i) Young exceptional performers and later adult world-class performers are largely two discrete populations over time. (ii) Early (e.g., youth) exceptional performance is associated with extensive discipline-specific practice, little or no multidisciplinary practice, and fast early progress. (iii) By contrast, adult world-class performance is associated with limited discipline-specific practice, increased multidisciplinary practice, and gradual early progress. These discoveries advance understanding of the development of the highest echelons of human achievement.

No ratings yet

View paper →

Your personalresearch library

All Papers

Content moderation and creator incentives mechanism amidst controversial content surges

Lewis Hine Young Boy Coal Miner, 1909-13 | Weisman Art Museum

On argumentation in Old English philology, with particular reference to the editing and dating of Beowulf | Anglo-Saxon England | Cambridge Core

Fatal Business Planning Assumptions

Why immigration research is probably biased

Dermoscopic features of lip squamous cell carcinoma and actinic cheilitis: A systematic review and meta‐analysis

2025 LLM Year in Review | karpathy

is it such a terrible thing to look like i've lived?

Next-Embedding Prediction Makes Strong Vision Learners

Distributional AGI Safety

On Reading Proust’s In Search of Lost Time

Recent discoveries on the acquisition of the highest levels of human performance