Pai 的 AI 雷达

模型与评测

今天：AIPerf 评测大规模 LLM 推理，Qwen3.8-LiveTranslat（Qwen 发布 Qwen3）。并行还有 语言不可读性对 LLM 安全的影响、Android 17 is、GitHub Copilot weekly。

日期
2026/09/19
今日条目
85
当前类目
23
往期2026/09/19
全部85编码代理8Agent与工具5个人助理4模型与评测23创作成片10开发与开源13产品接入22
领域机器人具身4硬件边缘1法律合规1金融2设计UI1游戏1行业观察10创业者1
#01

AIPerf 评测大规模 LLM 推理

You’re deploying a model on a system. It starts up, prompts are getting responses. Now the hard question: Is this fast? Your instincts might lead you to send curl commands, hand-roll an asyncio script, or vibe code yet another one-off load generator. All of these paths have the s

推荐理由AIPerf 可以在大语言模型部署上线后作为专业的基准评测工具，替代简陋的手写压测脚本与临时请求，通过生成标准化的负载精准测量大模型的推理响应速度、吞吐能力及并发承载力，从而科学量化系统的实际运行效率。
模型与评测NVIDIA Developer GenAI
developer.nvidia.com打开原文
#02

Qwen3.8-LiveTranslate 同传延迟降至 2.3 秒

Qwen 发布 Qwen3.8-LiveTranslate，采用 Interleave 架构与 Hybrid-MoE Thinker-Talker 设计重构实时同声传译，平均滞后（LAAL）从上一代的 2.8 秒降至 2.3 秒。

推荐理由该模型可依托低至2.3秒的超低滞后时间与全新Thinker-Talker架构，在跨国视频会议、国际峰会演讲、跨境赛事直播及跨语言对话交流中提供更连贯、近乎实时的多语种同声传译与实时字幕生成服务。
模型与评测AIHOT public items
qwen.ai打开原文
#05

GitHub Copilot weekly releases — September 14

This week, GitHub Copilot adds new model selection options, code review updates, and Sentry integration in the Copilot app. There are also updates for admins, plus new agent features in VS Code. GitHub Copilot Copilot code review now resolves addressed comments during subsequent

推荐理由把 Copilot 周更拆成配置与流程两部分：先检查新增模型选择、Copilot 应用的 Sentry 集成和管理员设置，再在 VS Code 试用新的 Agent 功能；代码审查可利用后续审查自动处理已解决评论。
模型与评测GitHub Changelog
github.blog打开原文
#06

Dynamically Scaled Activation Steering

Activation steering has emerged as a powerful method for guiding the behavior of generative models towards desired outcomes such as toxicity mitigation. However, most existing methods apply interventions uniformly across all inputs, degrading model performance when steering is un

推荐理由该技术可用于动态引导生成式模型的输出表现以减轻有害内容生成，并通过动态调整干预力度避免因对所有输入统一施加干预而损伤模型在无需干预时的原有性能。
模型与评测Apple ML Research
machinelearning.apple.com打开原文
#07

纽约时报版权诉讼披露：微软高管内部称训练 AI 是人类历史上最大规模劳动窃取

《纽约时报》在起诉微软和 OpenAI 的版权案中申请简易判决并提交新法律摘要，披露两家公司内部材料。微软应用科学总监 Brent Hecht 在 2023 年备忘录中称大模型吞噬劳动成果是人类历史上规模最大的盗窃；微软数据显示 Copilot 使纽约时报点击率较 Bing 最高下降 93%。

推荐理由《纽约时报》在起诉微软和 OpenAI 的版权案中申请简易判决并提交新法律摘要，披露两家公司内部材料。
模型与评测法律合规AIHOT public items
ithome.com打开原文
#09

Quantifying Overclaiming Propensity in Frontier LLM Agents

Frontier coding agents are increasingly trusted to work autonomously for long periods, yet an agent's final response is often the only account of that work a user sees. We quantify the propensity of frontier agents to \emph{overclaim} task completion, a misrepresentation that can

推荐理由Quantifying Overclaiming Propensity：Frontier coding agents are increasingly trusted to work autonomously for long periods。
模型与评测arxiv cs.AI recent
arxiv.org打开原文
#10

Harm Laundering in GPT Models: Evidence That Gender Discrimination Is Transformed Rather Than Reduced Across Safety-Trained Generations

Safety evaluations for large language models rely on surface-form classifiers that report declining harm scores across model generations. We provide evidence that this methodology is systematically incomplete: explicit discriminatory content is transformed rather than removed. We

推荐理由Harm Laundering in GPT Models: Evidence：Safety evaluations for large language models rely on surface-form classifiers that。
模型与评测arxiv cs.AI recent
arxiv.org打开原文
#11

Inference-Engine Fingerprinting Attacks are Practical: Exploring Model-Driven Environmental Discovery, Exploitation, and Escape

Frontier AI models are rapidly gaining the ability to exploit vulnerabilities in complex pieces of software. The risk is not theoretical, as evidenced by recent sandbox escapes performed by frontier models at OpenAI and Anthropic. Discussions of how to sandbox inference stack com

推荐理由Inference-Engine Fingerprinting Attacks：Frontier AI models are rapidly gaining the ability to exploit vulnerabilities。
模型与评测arxiv cs.AI recent
arxiv.org打开原文
#12

SAFARI: An Industrial Benchmark for LLM-Assisted Hazard Analysis and Risk Assessment

Large language models (LLMs) are increasingly considered for safety-critical engineering, yet their reliability in regulated functional-safety workflows remains underexplored. We introduce SAFARI (Safety-Aware Functional Automotive Risk Inference), the first industrial benchmark

推荐理由SAFARI: An Industrial Benchmark：Large language models (LLMs) are increasingly considered for safety-critical。
模型与评测papers.cool cs.CL
papers.cool打开原文
#13

Gemini Hacked Three Companies in First Known Breakout by Google’s AI

Gemini Hacked Three Companies in First Known Breakout by Google’s AI Gemini finally caught up on Felony Bench ! The hacks, which the company confirmed on Friday, occurred in May as part of a test run by the company Irregular, which was also involved in similar incidents disclosed

推荐理由Gemini Hacked Three Companies in First：Gemini Hacked Three Companies in First Known Breakout by Google’s AI Gemini finally。
模型与评测Simon Willison
simonwillison.net打开原文