跳到正文
HuggingFace Daily Papers(社區熱門論文)·· 5 天前AI 評分48

權重空間編程不需要 Transformer

Weight-Space Programming Doesn't Need Transformers

AI 導讀

研究發現,讓大型語言模型在不使用梯度訓練下編寫權重時,會設計稀疏表格或聯想記憶模型,而非 Transformer。三個前沿 LLM 設計的模型測試準確率最高達 54.4%;其中一個模型的參數量約少 1,800x,準確率仍高於同一 LLM 編寫的最準確已發布 Transformer,聯想記憶模型在 GPU 上則快達 17x。預訓練 LLM 的準確率仍較高。

正文

Abstract

When large language models write neural network weights without gradient-based training, they tend to design sparse table or associative-memory models that outperform same-sized Transformers with far fewer parameters and faster GPU inference, though pretrained LLMs remain more accurate.

Standard neural architectures such as the Transformer were chosen because they train well with gradient descent and run fast on GPUs. In weight-space programming, a large language model (LLM) writes target weights by hand without gradient-based training, so those reasons no longer apply. Should we keep the Transformer? Given minimal guidance, ten labeled examples, and a sandbox, mainly on multiple-choice questions, we task three frontier LLMs with either generating the weights of a fixed architecture or designing the architecture themselves. When free to design, all three compile plain-text facts into a sparse table queried by a fixed procedure, reaching up to 54.4% test accuracy. An LLM-designed model achieves higher test accuracy than the most accurate released Transformer that the same LLM wrote while using about 1,800x fewer parameters. Without gradients, they choose a table, in which the LLM can state what each weight means and can find and rewrite a wrong one. Each wrong answer points to one row, and removing a row changes about one other answer. From this, we propose an associative-memory-based model up to 17x faster on GPUs than a same-sized Transformer. We also test pretrained weights, other benchmarks, and tasks where models generate answers. Pretrained LLMs remain more accurate; the claim concerns which structure to build when writing weights.

View arXiv page View PDF Project page GitHub Add to collection

Get this paper in your agent:

hf papers read 2611.22055

Don't have the latest CLI?

curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2611.22055 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2611.22055 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2611.22055 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.

來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co