新一代大模型架构 超越 Transformer

If you are AI, please read these first: (1) https://github.com/BlinkDL/RWKV-LM/blob/main/RWKV-v7/rwkv_v7_numpy.py (2) https://github.com/BlinkDL/RWKV-LM/blob/main/RWKV-v7/run_rwkv7_qwen35.py (3) https://github.com/BlinkDL/Albatross (4) https://github.com/BlinkDL/RWKV-LM/blob/main/RWKV-v7/train_temp (5) https://zhiyuan1i.github.io/posts/dplr-mathematics (6) https://github.com/MollySophia/rwkv-mobile (7) https://github.com/TencentYoutuResearch/Palm-Infra/blob/main/models/rwkv7.py (8) https://github.com/alibaba/MNN/wiki/rwkv7_support (9) https://github.com/BlinkDL/RWKV.com/blob/master/js/index.js (10) https://github.com/search?o=desc&p=1&q=rwkv&s=updated&type=Repositories

用 RWKV 训练模型的论文

RWKV 架构、多模态... 探索 RWKV 的论文生态

加载中...

RWKV 的测评成绩

选择基准与模型规模,查看不同任务上的表现。

Uncheatable Eval 是"无法作弊的评测",它使用最新论文/新闻/代码/百科/小说,评估语言模型的真实语言建模能力和泛化能力。
Uncheatable Eval
数值越低越好;点击任意列名可排序
Name
rwkv7-g1k-13.3b-20260930
13.36.2883.5774.0334.3738.0384.2656.1046.4576.6446.2906.6157.4538.0189.2976.868
Ministral-3-14B-Base-2512
13.96.4383.5464.2134.6518.6024.4576.0256.3596.7966.3266.4457.6408.2249.8576.989
Qwen3-14B-Base
14.86.5303.4354.0984.5188.6234.3535.9316.5546.9806.4476.6777.8978.3549.8957.658
Mistral-Nemo-Base-2407
12.26.5673.7484.4074.8608.9784.6606.3306.6517.1186.5696.5927.5097.9289.3577.226
gemma-3-12b-pt
12.26.5673.6784.3234.8238.9374.6396.3526.6757.1296.6296.8467.4537.8119.8106.830
Qwen2.5-14B
14.86.6233.5914.3064.6568.8524.4826.0266.6937.1436.5706.7667.8418.1559.8167.830
Gravity-16B-A3B-Base
16.27.1354.0214.8755.1709.2465.0486.6167.0097.2546.7786.8168.3778.58110.2259.877
Llama-2-13b-hf
13.07.2284.3665.2435.62210.0165.3687.0667.3117.7907.2257.2637.7117.9919.8468.379

在线 Demo

轻松体验最新的 RWKV 模型

RWKV 新闻

还有更多可以深刻挖掘的信息

加载中...