🤯牛哇!开源圈把 Jev 的「系统一」快思考范式复刻到这个程度了,SemIf 在 JevBench 上 74.7 分,几乎贴着官方 Jev 1.13.0 的…


This is a dev post classified by Jev as Other (a tool drop), kept by the Dev Radar because it carries real work, not commentary.
🤯牛哇!开源圈把 Jev 的「系统一」快思考范式复刻到这个程度了,SemIf 在 JevBench 上 74.7 分,几乎贴着官方 Jev 1.13.0 的 75.4,暂列开源第一! 本地一张 3090 就能跑通 Jev 式快速思考 核心亮点: 🔹冻结 Qwen3.5-4B,几乎没改结构,直接读选项 token 的 logits,一次前向就出 typed 概率,不生成 JSON、不修格式 🔹3090卡上 21 个判断约 1 秒,比生成同样 JSON 快 5 倍+;还能浏览器 WebGPU 跑 🔹开源、可复现、有完整 runner 和行级结果;没复现 Jev 的 RLCD,校准也没专门做,分数接近但作者自己也说分数≠真实表现。 关键步骤: 1️⃣把状态 + 运行时定义的问题 + 2–16 个选项塞进 prompt 2️⃣模型只做一次前向,不 decode 3️⃣只对固定大写选项 token(A/B/C…)的 logits 做 softmax 4️⃣共享前缀可并行/串行打一堆判断 仓库:http://github.com/TheoLeeCJ/SemIf 独立项目,和 TypeSafe/Jev 无关。 #Jev #SemIf #SystemOne #开源复刻 #Qwen #JevBench #LLM #AIAgent
Posted by SuSu_酥酥👅 (15.8k followers) 1 h ago · 17 likes · 1.4k views · view the original post on X. Kept by the Dev Radar as Other. Tools mentioned: semif.
More dev work like this
- Large Vision Model! #BigData #Analytics #DataScience #AI #MachineLearning #NLProc #LLM… — @gp_pulipaka
- Building a WhatsApp automation bot shouldn’t mean wiring every command into the core. — @DanKornas
- I have previously said this: You don't pick an inference engine first. You pick a… — @TheAhmadOsman
- Meet Qwen-Image-2.1, the most balanced and cost-effective image generation model in the… — @alibaba_cloud
- Jev pod tomorrow — @swyx
- 😂笑死!Jev 刚出一周,中文圈圈已经把命名权当地皮抢完了,Jev-Anything! — @NFT_Chen
- 2.78 万亿参数,单 CPU,8.24GB 内存。 — @bkdgiffug
- Maestro v2.3 is out! — @blizaine
Every post is read and classified by Jev (TypeSafe): what it is, which market it belongs to, and whether the link is a real tool. 18.9k posts from 4.9k X accounts over the last 21 days, 2.1k tools, 12 markets. Collected every 5 minutes, fully re-ranked every hour — last update 2026-09-21 06:12 UTC. Full method.