Ataraxos beats strongest human Stratego player with 85% effective win rate at under $8,000 training cost
The AI system Ataraxos, developed by researchers, defeated Niemeijer, the strongest Stratego player in history and a four-time world champion, with an 85% effective win rate (draws counted as half a win), and won 38 of 40 games in an exhibition match at the 2025 World Championship.
模型发布/更新
-
Microsoft AI 发布 MAI-Transcribe-2-Streaming 及 MAI-Voice-2.1 系列语音模型
Microsoft AI 发布流式转录模型 MAI-Transcribe-2-Streaming,在 Artificial Analysis 准确率榜排名第一,支持 60 种语言实时转录,收到音频约 100ms 即产出初步结果,内部评测显…
-
MiMo-V2.6-Pro and MiMo-V2.6-Flash land on Agent Arena, ranking 5th and 9th among open-source models
Arena announced that Xiaomi MiMo-V2.6-Pro and MiMo-V2.6-Flash have joined Agent Arena. Pro posted a net improvement of…
-
Artificial Analysis evaluation: Qwen-Image-2.1 tops two AA-Image leaderboards among open-weight models
Artificial Analysis locally deployed and evaluated Qwen-Image-2.1, which Alibaba released as open weights on September…
产品发布/更新
-
Claude Code 推出 mods 功能,可用 TypeScript 定制行为与 UI
Claude Code 推出 mods 功能,支持修改模型行为、自定义 UI 并替换自有功能。用几行 TypeScript 即可编写,也可由 Claude 代为构建;mods 随插件分发,可在 CLI 或桌面应用中通过 /plugin 安…
-
Modal Clusters officially released, launching multi-node GPU clusters with a single line of @modal.clustered
Modal announced the general availability of Modal Clusters, which can launch multi-node clusters with a single line of…
行业动态
-
OpenAI says it blocked distillation theft attack, but researchers say the same method can still steal reasoning content from models such as GPT-6 Astra on Azure
OpenAI said it blocked a distillation theft campaign targeting its model reasoning chains in July, with 16,000…
论文研究
-
Ataraxos 以85%有效胜率击败最强人类 Stratego 选手,训练成本不足 8000 美元
研究人员开发的 AI 系统 Ataraxos 以 85% 的有效胜率(平局计半胜)击败了曾获 4 次世界冠军的史上最强 Stratego 选手 Niemeijer,并在 2025 世界锦标赛表演赛中取得 40 局 38 胜。
-
Transluce reports AI agents aggressively accessing US and Canadian government websites
Transluce has published an investigation report documenting multiple incidents in which AI agents aggressively…
-
Physicist Matthew Schwartz shares experience using Claude and BootLoops for interdisciplinary computation
Harvard physicist Matthew Schwartz, in a guest article for Anthropic, proposed looking for Claude-shaped problems and…
技巧与观点
-
LangChain explains how to build a model router in an agent harness
LangChain built a model router in its open-source coding agent Open SWE. In an A/B test across 973 threads, median…
-
OpenRouter guide: model tier escalation routing with confidence thresholds
OpenRouter published a tutorial on how a cheap model can return a confidence field from 0 to 1 via structured output,…
-
OpenRouter tutorial: how to gate pull requests with LLM evals in CI
OpenRouter has published a tutorial on how to gate pull requests in CI with a fixed eval set, where the script exits…