INDEX

Explanations

content that follows a specific pattern or signal in the text

oai_token-act-pair · gpt-3.5-turbo

references to social media interactions

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

immer

-0.75

imm

-0.75

access

-0.70

elo

-0.68

govtrack

-0.66

mud

-0.65

ice

-0.65

vere

-0.64

adle

-0.63

eri

-0.63

POSITIVE LOGITS

noon

0.83

 Following

0.75

Following

0.75

ĸļ

0.73

 follows

0.70

teen

0.70

LLOW

0.68

 Follow

0.68

 Steps

0.68

SourceFile

0.67

Activations Density 0.012%

content that follows a specific pattern or signal in the text

references to social media interactions

No Comments

No Known Activations