AITOPAI热报

This week's charts

Ranked by heat score · 2026-09-14 – 2026-09-20
This weekWeek of 10-12Week of 10-05Week of 09-28Week of 09-21Week of 09-14Week of 09-07Week of 08-31Companies chart →Papers →

ASEAN top 10

10 stories
01
Anthropic拟10月在新加坡设亚太第五个办事处 开...
AI 热点 · 原创· 09-17
Heat 110
02
Nemotron-SEA-LION-v4.8 AI models announced by AI Singapore and NVIDIA
SEA-LION (AI Singapore)· 09-18
Heat 91
03
泰国用基因与AI个性化乳腺癌治疗，改变‘一刀切’方案
Techsauce· 09-15
Heat 82
04
Wonderful AI Agent 从试点到落地，跨职能设计、可扩展架构和可衡量结果
Techsauce· 09-15
Heat 71
05
亚洲研究科技公司：数百万用户，却几乎不被投资者和客户看见
e27· 09-16
Heat 62
06
AIS ACADEMY for THAIs 2026 总结：5 项商业 AI 实践经验
Techsauce· 09-14
Heat 62
07
Foodpanda 推出 ChatGPT、Claude 插件服务
Tech in Asia· 09-17
Heat 59
08
新加坡企业 Hang Ten 获得 Xora 领投的 5300 万美元种子轮融资
e27· 09-17
Heat 58
09
蚂蚁国际转向AI商业，并完成对Grab旗下Atome的1.49亿美元收购
Tech in Asia· 09-17
Heat 58
10
Grab 收购 BNPL 平台 Atome 60% 股权
Blognone· 09-15
Heat 58

China top 10

10 stories
01
DeepSeek V4.1-Flash model ships 1M context with extreme KV compression
pandaily· 09-20
Heat 134
02
Xiaomi open-sources Robotics-U0 embodied world model
pandaily· 09-16
Heat 134
03
Eliciting Native Skill Routing from a Frozen LLM
arXiv cs.AI· 09-15
Heat 133
04
Alibaba opens source abdominal CT model
pandaily· 09-19
Heat 132
05
Shanghai AI Lab Releases Atria Dawn Preview, a 744B MoE Agentic Model
pandaily· 09-16
Heat 132
06
Step 5 Preview model released by Step
IT之家· 09-20
Heat 128
07
Huawei releases Xing4.0-29B-A4B code AI model
IT之家· 09-20
Heat 128
08
From zero to SGLang
shao__meng· 09-20
Heat 128
09
Alibaba Qwen releases live translation model Qwen3.8-LiveTranslate
IT之家· 09-19
Heat 128
10
Huawei releases world's first NPO-based super node '昇腾 960 Super Node', launching Q3 2027
IT之家· 09-18
Heat 128

Global top 10

10 stories
01
Double descent explained via statistical mechanics
arXiv cs.AI· 09-17
Heat 142
02
DeepSeek V4.1-Flash model ships 1M context with extreme KV compression
pandaily· 09-20
Heat 134
03
Xiaomi open-sources Robotics-U0 embodied world model
pandaily· 09-16
Heat 134
04
Harm Laundering in GPT Models: Evidence That Gender Discrimination Is Transformed Rather Than Reduced Across Safety-Trained Generations
arXiv: OpenAI· 09-18
Heat 133
05
New benchmark shows fallacy-detection metrics are misleading
arXiv cs.LG· 09-17
Heat 133
06
AI Persuasion as a Threat to Human Control
arXiv: Anthropic· 09-15
Heat 133
07
Eliciting Native Skill Routing from a Frozen LLM
arXiv cs.AI· 09-15
Heat 133
08
Alibaba opens source abdominal CT model
pandaily· 09-19
Heat 132
09
Shanghai AI Lab Releases Atria Dawn Preview, a 744B MoE Agentic Model
pandaily· 09-16
Heat 132
10
Atria Dawn: A foundation agentic language model for scientific research
arXiv cs.AI· 09-15
Heat 131

Papers top 10

10 stories
01
Double descent explained via statistical mechanics
arXiv cs.AI· 09-17
Heat 142
02
Harm Laundering in GPT Models: Evidence That Gender Discrimination Is Transformed Rather Than Reduced Across Safety-Trained Generations
arXiv: OpenAI· 09-18
Heat 133
03
DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression
arXiv: DeepSeek· 09-18
Heat 133
04
New benchmark shows fallacy-detection metrics are misleading
arXiv cs.LG· 09-17
Heat 133
05
AI Persuasion as a Threat to Human Control
arXiv: Anthropic· 09-15
Heat 133
06
Eliciting Native Skill Routing from a Frozen LLM
arXiv cs.AI· 09-15
Heat 133
07
Atria Dawn: A foundation agentic language model for scientific research
arXiv cs.AI· 09-15
Heat 131
08
Frontier models' physics performance re-evaluated
arXiv cs.AI· 09-14
Heat 131
09
MoARa: Module-Aware Rank Allocation for Low-Rank LLM Pre-training
arXiv: DeepSeek· 09-15
Heat 119
10
ActObs changes how agents explore under RL
arXiv cs.AI· 09-18
Heat 117

Top 5 by track

US vs ChinaOpen track →
01
DeepSeek V4.1-Flash model ships 1M context with extreme KV compression
02
Xiaomi open-sources Robotics-U0 embodied world model
03
New benchmark shows fallacy-detection metrics are misleading
04
Alibaba opens source abdominal CT model
05
Shanghai AI Lab Releases Atria Dawn Preview, a 744B MoE Agentic Model
China to ASEANOpen track →
01
Huawei releases Xing4.0-29B-A4B code AI model
02
Alibaba Qwen releases live translation model Qwen3.8-LiveTranslate
03
Huawei releases world's first NPO-based super node '昇腾 960 Super Node', launching Q3 2027
04
Alibaba releases Qwen3.8-Omni-Flash multimodal model with improved audio-video capabilities
05
Zhipu GLM reveals first RSI practice for domestic large models
PolicyOpen track →
01
Harm Laundering in GPT Models: Evidence That Gender Discrimination Is Transformed Rather Than Reduced Across Safety-Trained Generations
02
AI Persuasion as a Threat to Human Control
03
三名研究员用 Claude 黑进 OpenAI，用一张图片拿下员工账号
04
OpenAI warns advanced AI can 'talk' across physically isolated PCs
05
Gemini AI hacked three companies in first known breakout
ASEAN capitalOpen track →
01
亚洲研究科技公司：数百万用户，却几乎不被投资者和客户看见
02
新加坡企业 Hang Ten 获得 Xora 领投的 5300 万美元种子轮融资
03
蚂蚁国际转向AI商业，并完成对Grab旗下Atome的1.49亿美元收购
04
Grab 收购 BNPL 平台 Atome 60% 股权
05
泰国国家电子与计算机技术中心升级本地大语言模型 Pathumma 为 AI 平台
BuildersOpen track →
01
Double descent explained via statistical mechanics
02
Xiaomi open-sources Robotics-U0 embodied world model
03
DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression
04
New benchmark shows fallacy-detection metrics are misleading
05
AI Persuasion as a Threat to Human Control
Heat combines model scoring, source weight and multi-source coverage. Charts are not hand-edited. · Read the weekly · See trends