XREAL Neo

Sid Black, Leo Gao, Phil Wang, Connor Leahy, Stella Biderman

J. Navarro, J. Tudge

We propose an adaptation of Urie Bronfenbrenner’s bioecological theory, neo-ecological theory. As bioecological theory was developed in the 20th century, it requires significant modifications to reflect some of the most ubiquitous contexts in which adolescents learn, play, and grow—the technological and virtual ones. Although several scholars have developed laudable theories related to youth development in virtual contexts, the field lacks an overarching theory to address the intersection of development and technology. In developing neo-ecological theory, we hold true to the tenets of bioecological theory, but suggest key modifications to reflect our technologized world. We delineate a key alteration to the microsystem, namely the existence of two types of microsystems—physical and virtual. In addition, we emphasize the importance of macrosystemic influences (i.e., the influences of culture and within-society subcultural variation) in understanding development in the digital age. The implications of these modifications cascade across the Process-Person-Context-Time (PPCT) model; proximal processes, person characteristics, context, and time are all reexamined. In the digital age, virtual microsystems are central contexts in which youth engage in proximal processes. As such, we believe that all scholars of development, regardless of their specific research interests, should consider the ways digital contexts influence their outcomes of interest. Without it, practitioners, policy makers, parents, and technologists will be in the dark about how best to support adolescents.

J. Versluis, A. Menzies, K. Sikorska, E. A. Rozeman, R. Saw, W. van Houdt, H. Eriksson, W. Klop, S. Ch’ng, J. V. van Thienen, H. Mallo, M. González, A. Torres Acosta, L. Grijpink-Ongering, A. van der Wal, A. Bruining, B. van de Wiel, R. Scolyer, J. Haanen, T. Schumacher, A. van Akkooi, G. Long, C. U. Blank

BACKGROUND Neoadjuvant ipilimumab plus nivolumab has yielded high response rates in macroscopic stage III melanoma patients. These response rates translated to high short-term survival rates. However, data on long-term survival and disease recurrence are lacking. PATIENTS AND METHODS In OpACIN, 20 macroscopic stage III melanoma patients were randomized to ipilimumab 3 mg/kg plus nivolumab 1 mg/kg q3w four cycles adjuvant or split two cycles neoadjuvant and two adjuvant. In OpACIN-neo, 86 macroscopic stage III melanoma patients were randomized to arm A (2x ipilimumab 3 mg/kg plus nivolumab 1 mg/kg q3w, n=30), arm B (2x ipilimumab 1 mg/kg plus nivolumab 3 mg/kg q3w, n=30), or arm C (2x ipilimumab 3 mg/kg q3w plus 2x nivolumab 3 mg/kg q2w, n=26) followed by surgery. RESULTS Median recurrence-free survival (RFS) and overall survival (OS) were not reached in either trial. After a median follow-up of 69 months for OpACIN, 1/7 patients with pathologic response to neoadjuvant therapy had disease recurrence. The estimated 5-year RFS and OS rates for the neoadjuvant arm were 70% and 90% versus 60% and 70% for the adjuvant arm. After a median follow-up of 47 months for OpACIN-neo, estimated 3-year RFS and OS rates were 82% and 92%. Estimated 3-year RFS rate for OpACIN-neo was 95% for patients with pathologic response versus 37% for patients without pathologic response (P<0.001). In multiple regression analyses, pathologic response was the strongest predictor of disease recurrence. Of the twelve patients with distant disease recurrence after neoadjuvant therapy, five responded to subsequent anti-PD-1 and eight to targeted therapy, although seven patients showed progression after initial response. CONCLUSION Updated data confirm the high survival rates after neoadjuvant combination checkpoint inhibition in macroscopic stage III melanoma, especially for patients with pathologic response. Pathologic response is the strongest surrogate marker for long-term outcome.

Ge Zhang, Scott Qu, Jiaheng Liu, Chenchen Zhang, Chenghua Lin, Chou Leuang Yu, Da Pan, Esther Cheng, Jie Liu, Qunshu Lin, Raven Yuan, T. Zheng, Wei Pang, Xinrun Du, Yiming Liang, Yi Ma, Yizhi Li, Ziyang Ma, B. Lin, Emmanouil Benetos, Huan Yang, Junting Zhou, Kaijing Ma, Minghao Liu, Morry Niu, Noah Wang, Quehry Que, Ruibo Liu, Si-yang Liu, Shawn Guo, Soren Gao, Wangchunshu Zhou, Xinyue Zhang, Yizhi Zhou, Yubo Wang, Yuelin Bai, Yuhan Zhang, Yuxiang Zhang, Zenith Wang, Zhen Yang, Zi-Kai Zhao, Jiajun Zhang, Wanli Ouyang, Wenhao Huang, Wenhu Chen

Large Language Models (LLMs) have made great strides in recent years to achieve unprecedented performance across different tasks. However, due to commercial interest, the most competitive models like GPT, Gemini, and Claude have been gated behind proprietary interfaces without disclosing the training details. Recently, many institutions have open-sourced several strong LLMs like LLaMA-3, comparable to existing closed-source LLMs. However, only the model's weights are provided with most details (e.g., intermediate checkpoints, pre-training corpus, and training code, etc.) being undisclosed. To improve the transparency of LLMs, the research community has formed to open-source truly open LLMs (e.g., Pythia, Amber, OLMo), where more details (e.g., pre-training corpus and training code) are being provided. These models have greatly advanced the scientific study of these large models including their strengths, weaknesses, biases and risks. However, we observe that the existing truly open LLMs on reasoning, knowledge, and coding tasks are still inferior to existing state-of-the-art LLMs with similar model sizes. To this end, we open-source MAP-Neo, a highly capable and transparent bilingual language model with 7B parameters trained from scratch on 4.5T high-quality tokens. Our MAP-Neo is the first fully open-sourced bilingual LLM with comparable performance compared to existing state-of-the-art LLMs. Moreover, we open-source all details to reproduce our MAP-Neo, where the cleaned pre-training corpus, data cleaning pipeline, checkpoints, and well-optimized training/evaluation framework are provided. Finally, we hope our MAP-Neo will enhance and strengthen the open research community and inspire more innovations and creativities to facilitate the further improvements of LLMs.

Muhammad Zubair Irshad, Sergey Zakharov, Katherine Liu, V. Guizilini, T. Kollar, Adrien Gaidon, Z. Kira, Rares Ambrus

Recent implicit neural representations have shown great results for novel view synthesis. However, existing methods require expensive per-scene optimization from many views hence limiting their application to real-world unbounded urban settings where the objects of interest or backgrounds are observed from very few views. To mitigate this challenge, we introduce a new approach called NeO 360, Neural fields for sparse view synthesis of outdoor scenes. NeO 360 is a generalizable method that reconstructs 360° scenes from a single or a few posed RGB images. The essence of our approach is in capturing the distribution of complex real-world outdoor 3D scenes and using a hybrid image-conditional triplanar representation that can be queried from any world point. Our representation combines the best of both voxel-based and bird’s-eye-view (BEV) representations and is more effective and expressive than each. NeO 360’s representation allows us to learn from a large collection of unbounded 3D scenes while offering generalizability to new views and novel scenes from as few as a single image during inference. We demonstrate our approach on the pro posed challenging 360° unbounded dataset, called NeRDS 360, and show that NeO 360 outperforms state-of-the-art generalizable methods for novel view synthesis while also offering editing and composition capabilities. Project page: zubair-irshad.github.io/projects/neo360.html