[KongchangAI]
Model自然语言自编码器 / Natural Language Autoencoder

NLA

A class of meta-models that map the internal activations of a language model to natural-language explanations. The generated explanations are natural-language descriptions, which cannot be directly verified through forward propagation.

Timeline (last 90 days)

Nov 6

On held-out activation directions, MAEM's inversion performance is better than corpus search and the natural language autoencoder (NLA).

Unverified50%
Nov 6

In identifying the concepts behind steering vectors, MAEM's performance is roughly on par with NLA.

Unverified50%
Nov 6

Unlike the explanations generated by NLA, which are difficult to verify directly, the text examples generated by MAEM allow activation similarity to be computed via a single forward pass through the subject model, thereby yielding an objective verification score.

Unverified50%
Nov 3

A natural language autoencoder (NLA) is a type of model that maps a model's internal activation vectors to natural language explanations and can reconstruct the activation back from that explanation.

Unverified50%
Nov 3

NLA training uses an "explain-reconstruct" loop objective: given the activation of a certain layer, generate a textual description, then reconstruct the original activation from this text alone, optimizing the fidelity of the explanation with a reconstruction loss.

Unverified50%
Nov 3

The core of VERS is to use a natural language autoencoder (NLA) to produce textual explanations of and reconstruct model activations, and to steer the reasoning process by editing the explanations.

Unverified50%

All Facts (6)

Source Articles