Original Summary

Hi HN, we built TOD, a model and API for jevlike LLM pipelines.<p>we have people using it for, routing a ticket, classifying intent, deciding whether something is urgent, or rating severity.<p>Architecture - 12B Gemma fine-tuned to score options. since it is gemma based it has multimodality built in - When there are &gt;120 labels, we have a 150M retriever that narrows them to a shortlist (recall@16 is 0.9–1.0 on our evals), and the 12B model scores that shortlist. - Context goes up to 49k tokens, with up to 8 images per request.<p>we have open sourced it,<p>model: <a href="https:&#x2F;&#x2F;huggingface.co&#x2F;parsecai&#x2F;tod" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;parsecai&#x2F;tod</a> API: <a href="https:&#x2F;&#x2F;parseclab.ai&#x2F;tod" rel="nofollow">https:&#x2F;&#x2F;parseclab.ai&#x2F;tod</a><p>We are working on getting the RL environments to finetune for custom use cases.<p>would love to hear your opinions on this.

中文概览

中文标题： Show HN：TOD——基于任务导向设计的通用决策模型

介绍TOD：面向类LLM管线的决策模型与API，基于12B Gemma微调对选项打分；标签超过120个时先用150M检索器缩小候选范围，再由此模型打分；上下文最长49k token，每请求最多8张图片；模型与API已开源，并正开发用于自定义场景微调的RL环境。

  • 情报分类：开源项目与落地
  • 分类依据：发布并开源了模型与API的落地项目
  • 信息来源：Hacker News 新项目
  • 发布时间：2026/10/7 01:07:37