#Zirui
LightNews LightNews
larsmuckli.bsky.social
Title: Memory-based and surround-based pattern completion show distinct laminar profiles in human V1
@zirui-zhang.bsky.social, Clement Abbatecola, Y Huang, Y Lazarova, AT Paton, @marwimber.bsky.social, @chrislevelt.bsky.social, @mattlark.bsky.social, @larsmuckli.bsky.social & Lucy S Petro
October 23, 2026 at 10:35 AM
mathnt-bot.bsky.social
Zirui Jiang: Metric Diophantine approximation with primes on irrational lines https://arxiv.org/abs/2610.36380 https://arxiv.org/pdf/2610.36380 https://arxiv.org/html/2610.36380
October 28, 2026 at 6:49 AM
csai-bot.bsky.social
Zirui Ding, Xinxin Fan, Hao Yuan, Shaoye Luo, Men Niu, Yunfeng Lu: Uncovering Why LLMs Fail: From Internal Mechanisms to Adversarial Reasoning Trajectories https://arxiv.org/abs/2610.37369 https://arxiv.org/pdf/2610.37369 https://arxiv.org/html/2610.37369
October 28, 2026 at 6:40 AM
csne-bot.bsky.social
Cheikh Ahmed, Zhiyu Zhang, Kun Mao, Yong Zhang, Zirui Zhou: Organon : From Evolutionary Search to Recursive Self-Improving Agents for Scientific Discovery https://arxiv.org/abs/2610.33046 https://arxiv.org/pdf/2610.33046 https://arxiv.org/html/2610.33046
October 27, 2026 at 6:43 AM
csir-bot.bsky.social
Peichun Hua, He Sun, Zirui Zhuang, Qi Qi, Yiming Qiu, Shinan Liu, Yunming Xiao: Tessera: Structure-Completed Evidence Units for Web Retrieval-Augmented Generation https://arxiv.org/abs/2610.37080 https://arxiv.org/pdf/2610.37080 https://arxiv.org/html/2610.37080
October 28, 2026 at 6:42 AM
newsen.bsky.social
FAIR Health's New Board Members: Lynn Barr and Zirui Song Join Forces to Enhance Healthcare Mission#United_States#New_York#FAIR_Health#Lynn_Barr#Zirui_Song
FAIR Health's New Board Members: Lynn Barr and Zirui Song Join Forces to Enhance Healthcare Mission
Lynn Barr and Zirui Song have been elected to the Board of FAIR Health, bringing extensive experience in healthcare reform and public policy. Their leadership aims to enhance the organization's mission in promoting healthcare quality.
third-news.com
October 21, 2026 at 12:47 PM
arxiv-daily-bot.bsky.social
Uncovering Why LLMs Fail: From Internal Mechanisms to Adversarial Reasoning Trajectories

Zirui Ding (Institute of Computing Technology et al.

#arXiv #cs.AI
Uncovering Why LLMs Fail: From Internal Mechanisms to Adversarial Reasoning Trajectories
Large language models (LLMs) are usually evaluated by final-output correctness or safety, yet an endpoint label does not reveal how a failure develops throughout the reasoning process. We therefore ask whether ordinary reasoning failures and adversarially induced failures can be characterized using…
arxiv.org
October 29, 2026 at 2:21 AM
thedailynu.bsky.social
Residents pushed to protect Evanston Dog Beach during a “Bark-In” protest before Monday’s City Council meeting.
—
Reported by @zkahn.bsky.social and Caroline Zirui Zeng https://dailynorthwestern.com/2026/10/28/city/bark-in-protest-criticizes-impending-closure-of-evanston-dog-beach/
‘Bark-In’ protest criticizes impending closure of Evanston Dog Beach
More than one hundred two-legged and four-legged protesters traded the sandy shore of the Evanston Dog Beach for the sidewalk outside Lorraine H. M...
dailynorthwestern.com
October 28, 2026 at 4:01 PM
si-cv-graphics.bsky.social
𝗣𝗲𝗿𝗰𝗲𝗶𝘃𝗲, 𝗣𝗿𝗲𝗱𝗶𝗰𝘁, 𝗮𝗻𝗱 𝗣𝗹𝗮𝗻: 𝗮 𝗨𝗻𝗶𝗳𝗶𝗲𝗱 𝗪𝗼𝗿𝗹𝗱 𝗔𝗰𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝗲𝗹 𝗳𝗼𝗿 𝗘𝗴𝗼𝗰𝗲𝗻𝘁𝗿𝗶𝗰 𝗗𝗲𝘅𝘁𝗲𝗿𝗼𝘂𝘀 𝗠𝗮𝗻𝗶𝗽𝘂𝗹𝗮𝘁𝗶𝗼𝗻
Zirui Wu, Chen Guo, Tianjian Jiang ... Jie Song
arxiv.org/abs/2610.22363
Trending on scholar-inbox.com
October 19, 2026 at 6:00 AM
cscl-bot.bsky.social
Peter Cho-Ho Lam, Ziyi Wang, Zirui Zhou: Prefix-Tuning for Arbitrary Output Sequences on Pretrained Transformers https://arxiv.org/abs/2610.24123 https://arxiv.org/pdf/2610.24123 https://arxiv.org/html/2610.24123
October 21, 2026 at 6:39 AM
csce-bot.bsky.social
Zirui Wen, Shihao Yang, Feng Liu: Observability-Aligned Diffusion and Null-Space Completion for EEG Source Localization https://arxiv.org/abs/2610.27461 https://arxiv.org/pdf/2610.27461 https://arxiv.org/html/2610.27461
October 21, 2026 at 6:39 AM
cscl-bot.bsky.social
Peter Cho-Ho Lam, Ziyi Wang, Zirui Zhou: Prefix-Tuning for Arbitrary Output Sequences on Pretrained Transformers https://arxiv.org/abs/2610.24123 https://arxiv.org/pdf/2610.24123 https://arxiv.org/html/2610.24123
October 20, 2026 at 6:39 AM
cscv-bot.bsky.social
Wenhao Zhang, Zhongliang Zhou, Yi Ping, Zirui Wang, Sheng Li: What Language Adds, and at What Cost: Four Controlled Probes of Pathology Vision-Language Models https://arxiv.org/abs/2610.26805 https://arxiv.org/pdf/2610.26805 https://arxiv.org/html/2610.26805
October 20, 2026 at 6:44 AM
paper.bsky.social
[30/30] 278 Upvotes, 2 Comments, 2 Posts, arXiv:2610.11959

🆕MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement

Core Team, Zongming Qiao, Ziyue Hua, Zirui Ou, Zihao Yue, Zihan Jiang
Reinforcement learning (RL) is the central training paradigm for advancing large foundation models towards self-improvement. This report introduces the MiMo-V2.6 series, an omni-modal family that pushes the frontier of model intelligence by scaling RL compute. Prior to RL, we conduct mid-training on a broad multimodal corpus to provide ample exploration space, and build a solid infrastructure on the pretrained hybrid-SWA architecture to support subsequent scale-up. We scale RL compute along three dimensions: (1) larger batches and higher throughput, with an asynchronous training that consumes 1,568 samples and 2.7-3.7B tokens per step at context lengths of up to 1M; (2) more diverse and complex environments, spanning code, general, visual, and cyber domains under a mixture of agent harnesses; and (3) more grader compute, via groupwise agentic grading that yields more accurate reward signals for long-horizon tasks and steers the model towards shorter, more token-efficient solutions. To keep training stable at scale, we freeze the MoE router and establish a multi-layer defense against reward hacking. We further build infrastructure for mixed-task agentic RL, including a unified trajectory representation, high-concurrency multi-framework rollout, decoupled control and data planes, and training-inference consistency. We open-source the training dynamics, RL environments, and RL framework to facilitate reproduction and further research on scaled RL and model self-improvement.
October 22, 2026 at 12:05 AM
astroarxiv.bsky.social
Secular Evolution of Quasiperiodic Eruptions: eRO-QPE2. Zirui Zhang et. al. https://arxiv.org/abs/2610.21844
Figure 1 Figure 2 Figure 3 Figure 4
October 16, 2026 at 11:48 PM
arxiv-cs-cl.bsky.social
Soumye Singhal, Joseph Jennings, Ka Shun Shum, Alexander Bukharin, Khushi Bhardwaj, Zirui Liu, Venkat Srinivasan, Jonathan Li, Yash Jain, Kezhi Kong, Markus Kliegl, ...
Priming Base Models for Agentic Post-Training: An Open Midtraining Recipe
https://arxiv.org/abs/2610.22019
October 17, 2026 at 8:04 AM
krxiv-astro-ph.bsky.social
Secular Evolution of Quasiperiodic Eruptions: eRO-QPE2
https://arxiv.org/pdf/2610.21844
Zirui Zhang, Cong Zhou, Zhen Pan.
https://arxiv.org/abs/2610.21844
arXiv abstract link
arxiv.org
October 16, 2026 at 8:32 AM
astrophhe-bot.bsky.social
Zirui Zhang, Cong Zhou, Zhen Pan: Secular Evolution of Quasiperiodic Eruptions: eRO-QPE2 https://arxiv.org/abs/2610.21844 https://arxiv.org/pdf/2610.21844 https://arxiv.org/html/2610.21844
October 16, 2026 at 6:51 AM
arxiv-cs-cv.bsky.social
Zirui Wu, Chen Guo, Tianjian Jiang, Zeren Jiang, Zeming Li, Lingchen Yang, Longteng Duan, Ying Xue, Handi Yin, Shenyuan Gao, Martin R. Oswald, Jun Zhang, ...
Perceive, Predict, and Plan: a Unified World Action Model for Egocentric Dexterous Manipulation
https://arxiv.org/abs/2610.22363
October 16, 2026 at 8:28 PM
arxiv-cs-cl.bsky.social
Zixiang Xu, Zirui Song, Wanrong Zheng, Chiyu Zhang, Yue Huang, Yuyang Peng, Kexin Zhang, Jinkun Chen, Xiuying Chen, Xiangliang Zhang, Yan Liu, Laurent Itti, Frank Shyu, Xiaoqiang Luo, ...
Grid-Bench: Can LLMs Predict How Discrete Worlds Change?
https://arxiv.org/abs/2610.22312
October 16, 2026 at 4:07 PM
arxiv-daily-bot.bsky.social
Adaptive-Group Process Rewards for GUI Agents

Zirui Shang et al.

#arXiv #cs.AI
Adaptive-Group Process Rewards for GUI Agents
Training graphical user interface (GUI) agents with reinforcement learning suffers from sparse outcome rewards, leading to inefficient learning. Process rewards offer a promising direction for credit assignment to intermediate actions; however, accurately evaluating an action requires identifying e…
arxiv.org
October 16, 2026 at 2:00 AM
csai-bot.bsky.social
Zirui Shang, Yang Liu, Xin Shu, Zhi Gao, Xinxiao Wu, Lifeng Fan: Adaptive-Group Process Rewards for GUI Agents https://arxiv.org/abs/2610.17710 https://arxiv.org/pdf/2610.17710 https://arxiv.org/html/2610.17710
October 15, 2026 at 6:43 AM
eesssy-bot.bsky.social
Pranjal Sharma, Atharva Navsalkar, Zirui Xu, Vasileios Tzoumas: Optimistic Distributed Submodular Coordination with Applications to Multi-Target Tracking with Multiple Robots https://arxiv.org/abs/2610.19417 https://arxiv.org/pdf/2610.19417 https://arxiv.org/html/2610.19417
October 15, 2026 at 6:45 AM
arxiv-daily-bot.bsky.social
Adaptive-Group Process Rewards for GUI Agents

Zirui Shang et al.

#arXiv #cs.AI
Adaptive-Group Process Rewards for GUI Agents
Training graphical user interface (GUI) agents with reinforcement learning suffers from sparse outcome rewards, leading to inefficient learning. Process rewards offer a promising direction for credit assignment to intermediate actions; however, accurately evaluating an action requires identifying e…
arxiv.org
October 15, 2026 at 2:13 AM
cslg-bot.bsky.social
Zirui Peng, Heyi Zheng, Ziming Liu: Rethinking Activation Functions: Reducing Bias Can Improve Learning https://arxiv.org/abs/2610.14500 https://arxiv.org/pdf/2610.14500 https://arxiv.org/html/2610.14500
October 14, 2026 at 6:43 AM