Bilibili

bilibiliがQwen 3.5をベースに作った翻訳向けLLMのIndex-TranslateがCJKの翻訳に良いらしい。9Bモデルを使ってみたけど、わりといい感じ bilibili/Index-Translate: A Multilingual, Multimodal Translation Model Family: bilibili/Index-Translate

⭐ 1,7k звёзд: Bilibili выкатила Index-Translate — семейство мультиязычных мультимодальных моделей перевода. Наконец-то перевод, который видит картинки, а не только текст. Лови, пока хайп свежий: bilibili/Index-Translate

Index SLM Technical Report We present Index-1.9B, a series of open small language models developed at Bilibili. The series comprises four models: Index-1.9B-Base, a foundation model with 1.9 billion non-embedding parameters pre-trained on 2.8 trillion predominantly Chinese and English tokens; Index-1.9B-Pure, a control variant trained with an identical recipe but with all instruction-like data strictly filtered from the corpus; Index-1.9B-Chat, aligned from the base model with supervised fine-tuning and direct preference optimization; and Index-1.9B-Character, which augments the chat model with retrieval-augmented generation for few-shot role-playing customization. Pre-training employs a Warmup-Stable-Decay learning-rate schedule in which the concentration of curated data is raised substantially during the decay phase, together with a Norm-Head output layer that stabilizes training under large learning rates. On a suite of standard benchmarks covering examination, reasoning, mathematics, and code, Index-1.9B-Base attains an average score of 64.92, competitive with or exceeding open models of several times its size. We further report controlled studies on model depth, learning-rate magnitude and scheduling, the interaction between learning-rate decay and data quality, and the effect of including instruction data during pre-training, and we document an unexplained surge in benchmark performance midway through the constant-learning-rate phase. All models, together with evaluation code, are released at bilibili/Index-1.9B. Comments: 16 pages, 9 figures. v4: updated author list to add Yining Wang, Anzhen Li, and Jingying Liu

🤖 Index-Translate ⭐ 1,406 stars Text, audio, and video—translated seamlessly in any language. The next generation of open multimodal translation models has officially arrived. 🔗 bilibili/Index-Translate #AI #MachineLearning t.co

B站把翻译模型开源了，还白送 API。 · 图片翻译昨天刚上线：截图、表情包整张直接翻 · 语音直翻不换声：中文说，英日西语出 · 接沉浸式翻译插件，整页秒变双语 文字 150 语，图片语音长文档全包，2B 版笔记本能跑，1,326★。 bilibili/Index-Translate #AI翻译 #沉浸式翻译 #出海 #开源 t.co

开源了个翻译模型 Index-Translate 🌍 150 种语言，免费 API 已开，欢迎体验 大家好，我是 B 站 Index LLM 团队成员 最近开源了 Index-Translate ，基于 Qwen3.5 Base 做了翻译训练。支持 150 种文本语言，有 2B 、9B 和 35B-A3B preview 网页 demo 直接试：👉index-translate.bilibili.com 懒得下权重的话，我们现在开了免费 API ，兼容 OpenAI 接口。 改一下现有客户端的 base URL 就能接，也写了一个零外部依赖的 Python 调用脚本： 👉github.com/bilibili/Index-Translate/blob/main/inf… 除了翻译质量，也专门训练了指令遵循。可以塞术语表、指定文风、要求保留内容和格式。比如翻小说时固定人名和境界名，或者直接要求“别解释，只输出译文”。 顺便还做了几个分支： - 🎙️Echo：语音直接翻成字幕，或者带音色克隆的端到端语音翻译。 - ⏱️Homura：给译文设音节预算，适合要卡时长的场景。 - 📚Nailong：长文档翻译模型，减少分块造成的前后不一致。 Repo： bilibili/Index-Translate 权重： huggingface.co 反馈直接回帖或者提 issue 都行，谢谢。