INDEX

Explanations

instances of the letter 'e' followed by a number

oai_token-act-pair · gpt-3.5-turbo

references to electronic mail or emails

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 Kitchen

-0.83

ãĤ¤ãĥĪ

-0.82

 Katz

-0.75

 Painter

-0.71

 Anthem

-0.70

 Kats

-0.67

 Revel

-0.67

 guiActiveUnfocused

-0.64

DOWN

-0.64

 Heights

-0.63

POSITIVE LOGITS

tymology

1.31

gypt

1.15

ighth

1.12

agles

1.11

ugen

1.09

ternally

1.06

arthed

1.06

thereum

1.03

igen

1.02

ucl

0.97

Activations Density 0.021%

instances of the letter 'e' followed by a number

references to electronic mail or emails

No Comments

No Known Activations