
ALT First page of “TeleChat to Xing4: Advancing Frontier Agentic Models on Domestic Chips and Frameworks”, arXiv 2611.22115 (17 Nov 2026), by Xingchen AGI Lab, Baotian Hu, Bingkai Yang, Boyang Wang et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “We present Xing 4, a series of large language models that extends the TeleChat lineage through coordinated advances in model architecture and training infrastructure on Ascend.”

ALT First page of “K2-Horizon: A Frontier Fleet of Open Source Language Models”, arXiv 2611.21003 (16 Nov 2026), by K2 Team, Zhengzhong Liu, Haonan Li, Xuezhe Ma, Alexander Moreno et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “We present K2-Horizon, a fleet of six fully open language models, from a 0.9B dense model to a 375B-A23B MoE, built for general-purpose reasoning and agentic work across math and science, software engineering, deep research, office automation, and beyond.”

ALT First page of “Training Reasoning Models in Simulation to Natively Use Real Robots”, arXiv 2611.18562 (13 Nov 2026), by Jiaming Tang, Zhuoyang Zhang, Junxian Guo, Fangzhou Zhao, Yilong Zhao et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “Robot-use agents (RUAs) have emerged as a promising approach to general-purpose robot control.”

ALT First page of “ScaleVLA: Scaling Vision-Language-Action Models to 552 Billion Parameters”, arXiv 2611.21607 (16 Nov 2026), by Ruixi Yu, Dairu Liu, Xuchuan Chen, Zekun Qi, Zekai Li et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “Frontier Vision-Language Models (VLMs) such as GPT-6 Astra have demonstrated remarkable capabilities in robot control, suggesting that broad pretraining provides knowledge valuable for physical action.”

ALT First page of “Back to Pixels: Lifting Latent Video Diffusion into Pixel Space”, arXiv 2611.17336 (12 Nov 2026), by Yanke Song, Shuo Chen, Chris Rockwell, Guande He, Nan Liu et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “Frontier video generation models work in a VAE latent space that compresses video in both space and time.”

ALT First page of “MIRIX-2.0: Multi-Agent Memory System”, arXiv 2611.12068 (9 Nov 2026), by Wei-Chieh Huang, Weizhi Zhang, Jiayan Nan, Peng Xia and Yu Wang, showing the title, authors, abstract and the start of the paper. The abstract opens: “Long-term memory lets LLM agents personalize, stay consistent, and reuse what worked before.”

ALT First page of “MMFlash: Faster Speculative Decoding for Vision-Language Models”, arXiv 2611.18482 (13 Nov 2026), by Sicheng Feng, Zigeng Chen, Song Wang, Gongfan Fang and Xinchao Wang, showing the title, authors, abstract and the start of the paper. The abstract opens: “Speculative decoding accelerates generation by letting a lightweight drafter propose tokens for parallel verification by a target model.”

ALT First page of “Classical Energies Learn to Reason 1,000x Faster than Looped Models”, arXiv 2611.17309 (12 Nov 2026), by Benjamin Hoover, Hugh Blayney and James C. R. Whittington, showing the title, authors, abstract and the start of the paper. The abstract opens: “Looped models have emerged as a prominent paradigm for reasoning, with successes on simple reasoning puzzles such as Sudoku-Extreme and Maze-Unique.”

ALT A 3 by 3 grid of the first pages of nine arXiv papers, left to right, top to bottom: Energy-Based Reasoners: Your Recurrent Reasoner Should Be an Energy-Based Model; Embodied First Exam; Unbundling Multimodal Vision Encoding; Sage-Taiyi: A Metacognitive Architecture for Scientific Discovery; Evolving Infra: Agent-Adaptive LLM Serving at Scale; SwarmEvolve: Latent Agent Swarms for Evolutionary Discovery and Optimization; System Optimizations for Cost-Efficient Agent Harnesses; UGG: Unified Generative Geometry Model; SO-Arena: Benchmarking the Capability Spectrum and Reliability of System One Models.

ALT First page of “Review the Episodes, Refine the Lessons: Simple Memory Evolution for Embodied Agents”, arXiv 2611.11509 (8 Nov 2026), by Yifei Yao, Yude Zou, Xing Gao, Guanbo Jiao, Yukai Wang et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “Recent studies of embodied agents based on frontier multimodal large language models (MLLMs) develop harness engineering around the MLLM, such as harnesses with post-trained policies and recursive self-improvement loops.”

ALT First page of “UGG: Unified Generative Geometry Model”, arXiv 2611.14440 (10 Nov 2026), by Amani Kiruga, Zhiyi Li, Ruojin Cai, Hansen Lillemark, Xiaoshan Wu et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “Understanding 4D scenes is a core challenge in visual intelligence, requiring models to jointly capture scene geometry, appearance, and dynamics from partial observations.”

ALT First page of “An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale”, arXiv 2010.11929 (22 Oct 2020), by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited.”

ALT First page of “Unbundling Multimodal Vision Encoding”, arXiv 2611.17334 (12 Nov 2026), by Ellis Brown, Chris Hoang, Shengyi Qian, Shang-Wen Li, Mike Lewis et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “How multimodal language models should encode visual input remains up for debate.”

ALT First page of “Speculative Self-Distillation enables Efficient Knowledge Internalization”, arXiv 2611.12016 (9 Nov 2026), by Shayan Talaei, Agam Bhatia, Arshia Soltani Moakhar, Jonas Hübotter, Amin Saberi and Azalia Mirhoseini, showing the title, authors, abstract and the start of the paper. The abstract opens: “Many real-world applications require language models to internalize knowledge that was absent from pretraining, to incorporate continually updated facts, or to learn new domain skills.”

ALT First page of “The Surprising Effectiveness of Deleting Weights in LLM Post-Training”, arXiv 2611.10866 (7 Nov 2026), by Jack Lu, Zhenbang Yang, Mike Lasby, Tejas Pote, Yani Ioannou and Mengye Ren, showing the title, authors, abstract and the start of the paper. The abstract opens: “How little of a pretrained large language model (LLM) has to change for it to acquire new capabilities?”

ALT First page of “System Optimizations for Cost-Efficient Agent Harnesses”, arXiv 2611.12122 (9 Nov 2026), by Jongseok Park, Jiaxiang Yu, Eric Zhou, Tony Hong, Sunga Kim et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “AI agent harnesses play the critical role of providing LLMs with an interface to a computing environment.”

ALT First page of “Energy-Based Reasoners: Your Recurrent Reasoner Should Be an Energy-Based Model”, arXiv 2611.08977 (6 Nov 2026), by William Xingxu Chen, Alexi Gladstone, Zeyu Michael Li, Heng Ji and Yilun Du, showing the title, authors, abstract and the start of the paper. The abstract opens: “Looping is a powerful inductive bias for reasoning.”

ALT First page of “Evolving Infra: Agent-Adaptive LLM Serving at Scale”, arXiv 2611.12304 (9 Nov 2026), by Hao Kang, Weili Xu, Ziyang Li, Zheng Du, Hanchen Li et al., showing the title, authors, abstract and the start of the paper. The abstract opens: “Agentic workloads are a major source of LLM inference demand.”

ALT First page of “Don't Just Trust Your Memory: Evidence-Linked Memory Reduces Memory-Induced Degradation in LLM Agents”, arXiv 2611.14320 (10 Nov 2026), by Yukyeong Lee, Yumin Choi, Jinheon Baek and Sung Ju Hwang, showing the title, authors, abstract and the start of the paper. The abstract opens: “External memory enables LLM agents to improve from experience by distilling trajectories of past tasks into concise lessons for new tasks.”
