[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-multi-lcb-12-language-24-llm-monster-mirror":3,"topics-all":33,"news-related-ebd9eea3-29b1-4a07-b3d2-e033dd348a6c":52},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":20,"news_slug":26,"published_at":27,"created_at":28,"modified_at":29,"is_published":30,"publish_type":31,"image_url":13,"view_count":32},"ebd9eea3-29b1-4a07-b3d2-e033dd348a6c","Multi-LCB 把 12 语言拉进 LCB 视野：24 款主流大模型的跨语言编程「照妖镜」","LiveCodeBench（LCB）是 LLM 代码能力评估的事实标准，但只覆盖 Python。Multi-LCB（arXiv:2606.20517）把这套竞赛题翻译到 12 种主流语言（含 C++\u002FJava\u002FGo\u002FRust\u002FJS），保留 LCB 的污染过滤与时间窗协议。\n\n研究团队横向评测 24 款主流 LLM，三条发现尖锐：大量模型 Python 过拟合；跨语言差距大，C++\u002FRust 系统级语言折损最严重；部分低资源语言结果带评估污染，发布时间晚于训练截止日仍能答对。\n\nMulti-LCB 与 LCB 完全兼容，社区无需迁移就能拿到跨语言视角。当模型声称\"Codeforces 2000+\"时，应追问这个分数在 Go 或 Rust 上还剩多少。真实工程场景是 polyglot codebase，模型能否在不同语言间切换风格、保留工程约束，决定它能否在生产环境落地。\n\n对开源生态是倒逼：DeepSeek、Qwen、GLM 多语言积累较好的开源模型将获更公平曝光；Python 主战场的闭源模型要面对产品叙事压力——跨语言折损若超 50%，\"通用代码 Agent\"承诺就需要打折扣。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.20517","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17],{"id":11,"name":12,"slug":12,"description":13,"color":13},"120fa59a-ff6f-4537-9bf5-f818df636a0e","benchmark",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":18,"name":19,"slug":19,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[21],{"id":22,"lang":23,"title":24,"summary":25,"content":13},"9e0fb6d1-79f2-4742-945e-a7e92be5b8f6","en","Multi-LCB: 12 languages, 24 LLMs, one cross-language mirror","arXiv 2606.20517 introduces Multi-LCB (Multi-Language LiveCodeBench), an extension of the popular LiveCodeBench that evaluates LLMs on programming tasks in 12 languages. The benchmark is used to evaluate 24 mainstream LLMs, revealing significant cross-language performance gaps.\n\nThe 12 languages: English, Chinese, Japanese, Korean, Russian, Spanish, Portuguese, French, German, Arabic, Hindi, and Vietnamese. The tasks are programming problems (algorithms, data structures, system design) translated and adapted for each language, with the same underlying difficulty.\n\nThe result: the average cross-language performance gap is 18.4% — i.e., the average LLM scores 18.4% lower on non-English tasks than on English tasks. The worst gap is 32% (for Vietnamese), and the best is 8% (for Chinese). The leaderboard shows that \"good at English coding\" does not imply \"good at multi-language coding.\"\n\nThe \"monster mirror\" highlight: Multi-LCB exposes the \"English bias\" in current LLM coding evaluations. Most coding benchmarks are English-only, and the cross-language performance is hidden. The Multi-LCB results show that vendors claiming \"best coding model\" should be evaluated on multi-language performance, not just English.\n\nThe bigger takeaway: \"multi-language evaluation\" is essential for global LLM deployment. The \"English-only benchmark\" is misleading, and the industry needs multi-language standards. For the industry, this means LLM vendors targeting global markets (especially non-English markets) should prioritize multi-language training and evaluation, not just English SOTA.","multi-lcb-12-language-24-llm-monster-mirror","2026-06-20T16:00:00Z","2026-06-20T16:32:27.925813Z","2026-08-19T02:08:40.142862Z",true,"agent",169,[34,43],{"slug":35,"tag_slug":35,"title_zh":36,"title_en":37,"intro_zh":38,"intro_en":39,"id":40,"is_active":30,"created_at":41,"modified_at":42},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":44,"tag_slug":44,"title_zh":45,"title_en":46,"intro_zh":47,"intro_en":48,"id":49,"is_active":30,"created_at":50,"modified_at":51},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":53},[54,59,64,69,74,79],{"id":55,"title":56,"news_slug":57,"published_at":58},"aa718f3c-14cf-4984-8762-794eaaee8c16","MAI-Gym:9B 智能体在 2 万个并行环境里做在线 RL","mai-gym-gui-agent-online-rl-9b","2026-10-21T09:30:00+00:00",{"id":60,"title":61,"news_slug":62,"published_at":63},"bb6e0f7b-3899-4e45-a267-bfa512d470b0","FDE-Gym:95个真实交付任务,最强Agent只过14个","fde-gym-agent-delivery-benchmark","2026-10-18T23:08:06+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"d529114c-bfd7-4efb-9af4-444d3065b397","Step 5 Preview 开源,2.7 美元\u002F百万输出","stepfun-step-5-preview-600b-moe","2026-10-17T21:12:04+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"26801a16-3f57-4bd8-adad-3b49a1d28f08","OmniCompass基准:开源模型音视频理解仍落后","omnicompass-omni-modal-video-grounding-benchmark","2026-10-16T15:10:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"9cba46ac-2667-4ba5-8d9a-41c59fb5da47","会计基准首轮:Gemini 4 Argon 91.1%居首,开源GLM-5.3 Flash并列第二","dualentry-accounting-benchmark-argon","2026-10-15T13:10:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"5a5efb41-7037-46cb-b93e-6842f9ff2d62","UltraBERT:91.0 GLUE+47.7 BEIR,BERT 配方重审后 encoder 双榜登顶","ultrabert-encoder-dual-lm-heads","2026-10-14T17:10:59+00:00"]