# 冻结主干只训投影层：新模态不损原能力

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-08-25 12:20
- AIHOT 分数：42
- AIHOT 链接：https://aihot.virxact.com/items/cmt86pmya2t5mro7312phjr8q
- 原文链接：https://x.com/rohanpaul_ai/status/2092104681963360739

## AI 摘要

新研究提出“Projector Is All You Train”：为语言模型添加新模态只需训练连接编码器与主干的投影层，无需微调模型本身。3D 测试中，冻结主干的模型性能持平或超越联合微调，训练速度快一倍，而联合微调的 Llama 在 GSM8K 上从 86.96% 暴跌至 0.61%。

## 正文

You can teach a language model a new modality without ever touching the backbone's weights.

Train only the projector between encoder and backbone, and the capabilities you already shipped cannot regress.

To add a new modality to a language model, you only need to train the small projector that maps the encoder's output into the model's embedding space — fine-tuning the language model itself adds nothing reliable and wrecks its existing skills.

In their 3D tests, frozen-backbone models matched or beat jointly fine-tuned ones while training twice as fast, and the jointly fine-tuned Llama backbone collapsed from 86.96% to 0.61% on GSM8K.

– arxiv. org/abs/2608.19726

Title: "Projector Is All You Train"