INDEX

Explanations

phrases related to decisive actions or announcements

oai_token-act-pair · gpt-3.5-turbo

phrases indicating disapproval or unacceptable actions

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

plet

-0.70

 freeze

-0.68

 lifes

-0.63

 medicine

-0.62

 connector

-0.60

 bureau

-0.60

zip

-0.60

 ultraviolet

-0.59

 literacy

-0.59

 VIDE

-0.58

POSITIVE LOGITS

 Whilst

1.37

 Needless

1.12

Upon

1.05

Following

1.05

Nevertheless

1.04

Thankfully

1.04

However

1.02

Furthermore

1.01

Nonetheless

1.00

 Thankfully

1.00

Activations Density 1.045%

phrases related to decisive actions or announcements

phrases indicating disapproval or unacceptable actions

No Comments

No Known Activations