#binyuan
LightNews LightNews
ragerat.bsky.social
#svsss #binqiu #binyuan
January 12, 2025 at 10:49 AM
krupamannka.bsky.social
Неуклюжая открытка на новый год для подруги
До сих пор в шоке от того, какой милоый она получилась

#SVSSS #ShenQingqiu #LuoBinghe #ShenYuan #BingQiu #BinQiu #BingYuan #BinYuan
January 20, 2025 at 12:34 PM
neurasensory.bsky.social
Li Binyuan: The Performance Art of Risk, Absurdity, and Rebellion - radii.co

#art
radii.co
news.google.com
August 18, 2025 at 5:45 AM
esrig.bsky.social
Congratulations to Dr. Binyuan Liu on successfully defending his PhD dissertation:

“Multi-scale greenhouse gas emissions accounting in developing countries”

🔗 More info:
www.rug.nl/about-ug/lat...

#PhDDefense #ClimateAction #Sustainability
June 12, 2026 at 3:34 PM
arxiv-cs-cl.bsky.social
Niklas Muennighoff, Qian Liu, Armel Zebaze, Qinkai Zheng, Binyuan Hui, Terry Yue Zhuo, Swayam Singh, Xiangru Tang, Leandro von Werra, Shayne Longpre
OctoPack: Instruction Tuning Code Large Language Models
https://arxiv.org/abs/2308.07124
February 20, 2024 at 9:11 AM
france.theconversation.com
Marina Abramović, Li Binyuan et Paula Garcia… Quand les performances artistiques apprennent à observer le fonctionnement des organisations
Marina Abramović, Li Binyuan et Paula Garcia… Quand les performances artistiques apprennent à observer le fonctionnement des organisations
Même s’ils semblent opposés, le monde des performances artistiques et celui des managers ont tout à gagner à dialoguer. Pour ces derniers, c’est un moyen de réfléchir à leurs pratiques.
dlvr.it
April 14, 2026 at 3:21 PM
arxiv-cs-cl.bsky.social
Chengwen Qi, Ren Ma, Bowen Li, He Du, Binyuan Hui, Jinwang Wu, Yuanjun Laili, Conghui He
Large Language Models Meet Symbolic Provers for Logical Reasoning Evaluation
https://arxiv.org/abs/2502.06563
February 11, 2025 at 8:07 AM
arxiv-cs-cv.bsky.social
Zhihong Tan, Jiayi Wang, Huiying Shi, Binyuan Huang, Hongchen Wei, Zhenzhong Chen
RSFAKE-1M: A Large-Scale Dataset for Detecting Diffusion-Generated Remote Sensing Forgeries
https://arxiv.org/abs/2505.23283
May 30, 2025 at 8:44 AM
techimo.bsky.social
Alibaba Qwen Team's technical lead, Junyang Lin, and key member Binyuan Hui have resigned. Lin oversaw releases like Qwen 1 & 3.5. Rumors cite unmet Daily Active User (DAU) targets as a potential reason for departures.
March 4, 2026 at 12:46 AM
ghastreblyansky.bsky.social
Too bad that's the task that LLMs are rock-bottom worst at, Elon.
Jinyang Li1,♣ ‡, Binyuan Hui2,♣, Ge Qu1,♣, Jiaxi Yang2 , Binhua Li2 , Bowen Li6 , Bailin Wang5 , Bowen Qin2 , Ruiying Geng2 , Nan Huo1 , Xuanhe Zhou3 , Chenhao Ma6 , Guoliang Li3 , Kevin C.C. Chang7 , Fei Huang2 , Reynold Cheng1 , Yongbin Li2 1 The University of Hong Kong 2 DAMO Academy, Alibaba Group 3 Tsinghua University 4 Shanghai AI Laboratory 5 MIT CSAIL 6 The Chinese University of Hong Kong, Shenzhen 7 University of Illinois at Urbana-Champaign {jl0725,quge}@connect.hku.hk, ckcheng@cs.hku.hk binyuan.hby@alibaba-inc.com Abstract Text-to-SQL parsing, which aims at converting natural language questions into executable SQLs, has gained increasing attention in recent years. In particular, GPT-4 and Claude-2 have shown impressive results in this task. However, most of the prevalent benchmarks, i.e., Spider, and WikiSQL, focus on database schema with few rows of database values leaving the gap between academic study and real-world applications. To mitigate this gap, we present BIRD, a B
April 1, 2024 at 3:27 PM
paper.bsky.social
[28/30] 113 Likes, 4 Comments, 2 Posts
2503.04625, cs․CL, 07 Mar 2025

🆕START: Self-taught Reasoner with Tools

Chengpeng Li, Mingfeng Xue, Zhenru Zhang, Jiaxi Yang, Beichen Zhang, Xiang Wang, Bowen Yu, Binyuan Hui, Junyang Lin, Dayiheng Liu
Large reasoning models (LRMs) like OpenAI-o1 and DeepSeek-R1 have demonstrated remarkable capabilities in complex reasoning tasks through the utilization of long Chain-of-thought (CoT). However, these models often suffer from hallucinations and inefficiencies due to their reliance solely on inte...
March 16, 2025 at 12:05 AM
android-adepts.bsky.social
Alibaba's Qwen AI team hit by major departures: Tech Lead Junyang Lin & key member Binyuan Hui have left. Lin led Qwen 1.0-3.5 releases. Reports suggest a Google Gemini team member may succeed Lin, and DAU metric failures could be the reason for the exits. #Alibaba #Qwen #AI
March 4, 2026 at 12:46 AM
arxiv-cs-cv.bsky.social
Chenyang Wu, Fuchen Long, Binyuan Huang, Xinlong Sun, Xi Chen, Chun-Le Guo, Chongyi Li
Thinking on Shots: Consistent Multi-Shot Video Editing with Agentic Reasoning
https://arxiv.org/abs/2608.26809
August 28, 2026 at 9:00 PM
cscv-bot.bsky.social
Chenyang Wu, Fuchen Long, Binyuan Huang, Xinlong Sun, Xi Chen, Chun-Le Guo, Chongyi Li: Thinking on Shots: Consistent Multi-Shot Video Editing with Agentic Reasoning https://arxiv.org/abs/2608.26809 https://arxiv.org/pdf/2608.26809 https://arxiv.org/html/2608.26809
August 28, 2026 at 6:40 AM
arxiv-cs-cl.bsky.social
Niklas Muennighoff, Zhengyang Wang, Zeyi Chen, Weijia Shi, Binyuan Hui, John Yang, Dapeng Jiang, Mika Senghaas, Fares Obeid, Johannes Hagemann, Sami Jaghouar, Ludwig Schmidt, Percy Liang, Jason Wei, ...
Prefix Sliding for efficient test-time scaling
https://arxiv.org/abs/2608.26070
August 27, 2026 at 2:43 PM
paper.bsky.social
[27/30] 292 Likes, 2 Comments, 2 Posts
2505.10527, cs․CL, 18 May 2025

🆕WorldPM: Scaling Human Preference Modeling

Binghai Wang, Runji Lin, Keming Lu, Le Yu, Zhenru Zhang, Fei Huang, Chujie Zheng, Kai Dang, Yang Fan, Xingzhang Ren, An Yang, Binyuan Hui, Dayiheng Liu, Tao Gui, Qi Zhang, Xuanjin...
Motivated by scaling laws in language modeling that demonstrate how test loss scales as a power law with model and dataset sizes, we find that similar laws exist in preference modeling. We propose World Preference Modeling$ (WorldPM) to emphasize this scaling potential, where World Preference embodies a unified representation of human preferences. In this paper, we collect preference data from public forums covering diverse user communities, and conduct extensive training using 15M-scale data across models ranging from 1.5B to 72B parameters. We observe distinct patterns across different evaluation metrics: (1) Adversarial metrics (ability to identify deceptive features) consistently scale up with increased training data and base model size; (2) Objective metrics (objective knowledge with well-defined answers) show emergent behavior in larger language models, highlighting WorldPM's scalability potential; (3) Subjective metrics (subjective preferences from a limited number of humans or AI) do not demonstrate scaling trends. Further experiments validate the effectiveness of WorldPM as a foundation for preference fine-tuning. Through evaluations on 7 benchmarks with 20 subtasks, we find that WorldPM broadly improves the generalization performance across human preference datasets of varying sizes (7K, 100K and 800K samples), with performance gains exceeding 5% on many key subtasks. Integrating WorldPM into our internal RLHF pipeline, we observe significant improvements on both in-house and public evaluation sets, with notable gains of 4% to 8% in our in-house evaluations.
June 3, 2025 at 12:07 AM
paper.bsky.social
[28/30] 385 Likes, 4 Comments, 3 Posts
2505.10475, cs․LG | cs․CL, 15 May 2025

🆕Parallel Scaling Law for Language Models

Mouxiang Chen, Binyuan Hui, Zeyu Cui, Jiaxi Yang, Dayiheng Liu, Jianling Sun, Junyang Lin, Zhongxin Liu
It is commonly believed that scaling language models should commit a significant space or time cost, by increasing the parameters (parameter scaling) or output tokens (inference-time scaling). We introduce the third and more inference-efficient scaling paradigm: increasing the model's parallel computation during both training and inference time. We apply $P$ diverse and learnable transformations to the input, execute forward passes of the model in parallel, and dynamically aggregate the $P$ outputs. This method, namely parallel scaling (ParScale), scales parallel computation by reusing existing parameters and can be applied to any model structure, optimization procedure, data, or task. We theoretically propose a new scaling law and validate it through large-scale pre-training, which shows that a model with $P$ parallel streams is similar to scaling the parameters by $O(\log P)$ while showing superior inference efficiency. For example, ParScale can use up to 22$\times$ less memory increase and 6$\times$ less latency increase compared to parameter scaling that achieves the same performance improvement. It can also recycle an off-the-shelf pre-trained model into a parallelly scaled one by post-training on a small amount of tokens, further reducing the training budget. The new scaling law we discovered potentially facilitates the deployment of more powerful models in low-resource scenarios, and provides an alternative perspective for the role of computation in machine learning.
May 31, 2025 at 2:50 AM
paper.bsky.social
[21/30] 134 Likes, 3 Comments, 1 Posts
2505.09388, cs․CL, 14 May 2025

🆕Qwen3 Technical Report

An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Hu...
In this work, we present Qwen3, the latest version of the Qwen model family. Qwen3 comprises a series of large language models (LLMs) designed to advance performance, efficiency, and multilingual capabilities. The Qwen3 series includes models of both dense and Mixture-of-Expert (MoE) architectures, with parameter scales ranging from 0.6 to 235 billion. A key innovation in Qwen3 is the integration of thinking mode (for complex, multi-step reasoning) and non-thinking mode (for rapid, context-driven responses) into a unified framework. This eliminates the need to switch between different models--such as chat-optimized models (e.g., GPT-4o) and dedicated reasoning models (e.g., QwQ-32B)--and enables dynamic mode switching based on user queries or chat templates. Meanwhile, Qwen3 introduces a thinking budget mechanism, allowing users to allocate computational resources adaptively during inference, thereby balancing latency and performance based on task complexity. Moreover, by leveraging the knowledge from the flagship models, we significantly reduce the computational resources required to build smaller-scale models, while ensuring their highly competitive performance. Empirical evaluations demonstrate that Qwen3 achieves state-of-the-art results across diverse benchmarks, including tasks in code generation, mathematical reasoning, agent tasks, etc., competitive against larger MoE models and proprietary models. Compared to its predecessor Qwen2.5, Qwen3 expands multilingual support from 29 to 119 languages and dialects, enhancing global accessibility through improved cross-lingual understanding and generation capabilities. To facilitate reproducibility and community-driven research and development, all Qwen3 models are publicly accessible under Apache 2.0.
May 21, 2025 at 12:07 AM
paper.bsky.social
[6/30] 234 Likes, 8 Comments, 1 Posts
2412.15115, cs․CL, 19 Dec 2024

🆕Qwen2.5 Technical Report

Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiax...
December 21, 2024 at 12:05 AM
paper.bsky.social
[27/30] 65 Likes, 2 Comments, 1 Posts
2409.12186, cs․CL, 18 Sep 2024

🆕Qwen2.5-Coder Technical Report

Binyuan Hui, Jian Yang, Zeyu Cui, Jiaxi Yang, Dayiheng Liu, Lei Zhang, Tianyu Liu, Jiajun Zhang, Bowen Yu, Kai Dang, An Yang, Rui Men, Fei Huang, Xingzhang Ren, Xuancheng Ren, Jingren Zhou, Ju...
September 20, 2024 at 12:05 AM
paper.bsky.social
[17/30] 92 Likes, 2 Comments, 2 Posts
2407.10671, cs․CL | cs․AI, 15 Jul 2024

🆕Qwen2 Technical Report

An Yang, Baosong Yang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Zhou, Chengpeng Li, Chengyuan Li, Dayiheng Liu, Fei Huang, Guanting Dong, Haoran Wei, Huan Lin, Jialong Tang, Jialin Wang, Jian Y...
July 17, 2024 at 12:05 AM
krxiv-cond-mat.bsky.social
Optically Driven Orbital Hall Transport in Floquet Odd-Parity Collinear Altermagnets with High Chern Numbers
https://arxiv.org/pdf/2603.11483
Yuping Tian, Chen-Hao Zhao, Chao-Bo Wang, Binyuan Zhang, Xiangru Kong, Wei-Jiang Gong.
https://arxiv.org/abs/2603.11483
arXiv abstract link
arxiv.org
March 13, 2026 at 4:34 AM
cscv-bot.bsky.social
Binyuan Huang, Yuning Lu, Weinan Jia, Hualiang Wang, Mu Liu, Daiqing Yang: Rethinking Position Embedding as a Context Controller for Multi-Reference and Multi-Shot Video Generation https://arxiv.org/abs/2604.03738 https://arxiv.org/pdf/2604.03738 https://arxiv.org/html/2604.03738
April 7, 2026 at 6:31 AM
cscv-bot.bsky.social
Weinan Jia, Yuning Lu, Mengqi Huang, Hualiang Wang, Binyuan Huang, Nan Chen, Mu Liu, Jidong Jiang, Zhendong Mao: MoGA: Mixture-of-Groups Attention for End-to-End Long Video Generation https://arxiv.org/abs/2510.18692 https://arxiv.org/pdf/2510.18692 https://arxiv.org/html/2510.18692
October 22, 2025 at 6:31 AM
cscv-bot.bsky.social
Binyuan Huang, Yongdong Luo, Xianda Guo, Xiawu Zheng, Zheng Zhu, Jiahui Pan, Chengju Zhou: Watch Where You Move: Region-aware Dynamic Aggregation and Excitation for Gait Recognition https://arxiv.org/abs/2510.16541 https://arxiv.org/pdf/2510.16541 https://arxiv.org/html/2510.16541
October 21, 2025 at 6:31 AM