INDEX

Explanations

references to buttons being clicked or pressed

oai_token-act-pair · gpt-3.5-turbo

instances of the word "button."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Flavoring

-0.74

ctuary

-0.72

 Continuing

-0.68

 Instit

-0.68

AST

-0.68

 Seraph

-0.68

ARS

-0.66

 Plants

-0.65

 Evening

-0.65

 Alleg

-0.65

POSITIVE LOGITS

button

1.09

 button

1.04

 buttons

0.97

bell

0.93

holes

0.91

pus

0.86

 clicked

0.86

oola

0.86

 Button

0.86

hole

0.86

Activations Density 0.015%

references to buttons being clicked or pressed

instances of the word "button."

No Comments

No Known Activations