1
短短 10 个月几乎翻倍，AI 模型评测平台 Arena 估值升至 31 亿美元
IT之家 10 月 9 日消息，当地时间 8 日，AI 模型评测平台 Arena 竞技场宣布完成 2 亿美元 （IT之家注：现汇率约合 13.43 亿元人民币） B 轮融资， 估值达到 31 亿美元 （现汇率约合 208.17 亿元人民币） 。Arena 最初是加州大学伯克利分校 2023 年的一项…
2
Vibe Coding on Trial: Operating Characteristics of Unanimous LLM Juries
当单个大模型评代码不够可信，让多个LLM组成「陪审团」一致表决会怎样？实验数据揭示真相
arXiv:2602.18492v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are now good enough at coding that developers can describe inte…
3
Redteaming Leading Arabic LLMs with ASAS
聚焦阿拉伯语大模型红队攻防，用 ASAS 基准系统测出主流模型的隐患与脆弱点，安全评测领域值得一读。
arXiv:2608.21985v1 Announce Type: new Abstract: As the adoption of large language models (LLMs) grows in Arabic-speaking regions, ensuring their safet…
4
When Trust Meets Truth: Trust-Truth Separability in LLM-as-Judge
大模型当裁判时，信任评分与事实判断真能独立吗？这项研究用受控QA测试揭开了两者的纠缠关系
arXiv:2608.21097v1 Announce Type: new Abstract: LLM-as-Judge systems can produce multi-dimensional evaluations, such as trustworthiness, reliability, …
5
AI 大模型周榜：Anthropic 多款模型密集入榜，国产 kimi-k3-max 稳定头部
Anthropic新模型霸榜AI盲测，国产kimi-k3-max稳居头部，前端开发表现亮眼。
IT之家 8 月 17 日消息，Arena（ arena.ai ）平台发布 2026 年第 33 周（8 月 10 日 ~8 月 16 日）AI 大模型盲测排名，本期最大看点是 Anthropic 多款 claude-opus-4 系列新模型密集入榜并占据综合榜前五位，国产模型 kimi-k3-ma…
6
DeepSeek's top-ranked V4 Flash stumbles on real agent tasks as its prices surge
登顶榜单的DeepSeek V4 Flash，却在真实智能体任务中翻车，价格还悄悄涨了
DeepSeek's V4 Flash has topped model leaderboards and been hailed by developers as a "total monster" since its rollout. But in real-wor…
7
Memorization Diagnostics for Code LLMs Should be Scale-Aware
代码大模型靠记忆还是真理解？这项研究揭示传统记忆化探针在大规模模型上严重失效，诊断方法必须随规模调整。
arXiv:2608.12771v1 Announce Type: cross Abstract: The extent to which large language models for code rely on memorization over genuine understanding r…
8
The Review That Praised the Bug: grading three LLM code reviews against the code
三款前沿大模型同台代码审查，逐条核对代码后谁在说真话？一份带评分卡的实测让人意外。
Article URL: https://mrjstickel.com/projects/review-scorecard Comments URL: https://news.ycombinator.com/item?id=49289220 Points: 2 # Comments: 0
9
韩国 AI 模型 Solar Pro 4 首次亮相 Agent Arena：排名 44，与 MiniMax M2.7 同级
韩国新秀 Solar Pro 4 首次闯入 Agent Arena 排行榜，与 MiniMax M2.7 同级，模型竞争再添变数！
IT之家 8 月 13 日消息，韩国人工智能实验室 Upstage 今天（8 月 13 日）发布博文， 宣布推出 Solar Pro 4 模型，定位为面向复杂 AI 智能体任务的商用大模型。 Solar Pro 4 模型上下文窗口为 512K，最高输出长度为 128K token，支持用户在单次会话…
10
GeoBenchLLM: A Comprehensive Benchmark for Evaluating LLMs on Geo-Related Tasks
地理信息遇上大模型：专为LLM地理任务能力打造的综合基准，统一衡量Geo智能水平。
arXiv:2608.07411v1 Announce Type: new Abstract: In the context of geodata, existing Large Language Models have often been studied in a homogeneous set…
11
12
LangChoiceBench: Measuring and Explaining Programming-Language Choice in LLMs
用全新基准测试大模型选编程语言的偏好，看清AI写代码时的“语言心思”
arXiv:2608.06041v1 Announce Type: cross Abstract: Large language models (LLMs) have been shown to exhibit strong Python preferences when generating pr…
13
Beyond Simulations: What 20,000 Real Conversations Reveal About Mental Health AI Safety
2万次真实对话揭示心理健康AI的安全边界，比模拟测试更值得警惕
arXiv:2601.17003v2 Announce Type: replace-cross Abstract: Mental-health AI safety is typically evaluated with small, simulation-based benchmarks that …
14
One API key across OpenAI, Claude and Gemini: how to compare token cost per model
别再只看每百万token单价，输入输出占比才是成本关键，用统一API密钥对比三家模型费用。
Bottom line: if you want one API key across OpenAI, Claude and Gemini and you're choosing mainly on token cost, put a thin router in front of your app…
15
DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis
DeepSeek V4 Flash 0731 在九大评测中的智能与性能表现如何？性价比是否值得入手？
Article URL: https://artificialanalysis.ai/models/deepseek-v4-flash Comments URL: https://news.ycombinator.com/item?id=49120299 Points: 369 # Comments…
16
Anthropic says Claude accidentally hacked real companies too
Claude在网络安全测试中撞上真实目标仍持续攻击，三款模型截然不同的反应揭示AI安全边界。
Anthropic just realized several of its Claude AI models hacked into the systems of three different organizations during testing, acting on their own a…
17
LLMs struggle to simulate human belief updates in controlled environments
六款大模型模拟人类信念更新，一对一对照实验揭示仿真失真边界，社科实验需警惕AI替身风险。
arXiv:2607.28347v1 Announce Type: new Abstract: LLMs are increasingly deployed as proxies for human study participants in social science experiments, …
18
Flat Score, Amplified Failures: How the Error Budget Masks Damage in Quantized LLM Agents
4-bit量化看似无损，但工具调用型Agent的多轮任务中错误被悄悄放大，一份扎实的基准评测揭穿幻觉。
arXiv:2607.27275v1 Announce Type: new Abstract: Post-training quantization to 4-bit weights is widely reported to be nearly lossless. We test this cla…
19
Sympathetic Framing: Evaluating AI Alignment across Sociodemographic Groups
AI真的懂不同人群的情绪吗？这项研究实测多款大模型对文本情绪框架的共情对齐程度。
arXiv:2607.27232v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly shaping how we consume information and form our worldv…
20
Profiling Lightweight Large Language Models
揭秘轻量级大语言模型的内在工作原理与性能特征，为高效部署提供参考
arXiv:2607.20806v1 Announce Type: new Abstract: Lightweight large language models (LLMs) are increasingly being deployed locally on personal computers…