INDEX

Explanations

words related to disguise or hiding

oai_token-act-pair · gpt-3.5-turbo

the word "gu" in various contexts

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

 Spectrum

-0.78

hower

-0.74

 Borders

-0.68

 closures

-0.68

heid

-0.68

Charges

-0.66

 cycle

-0.66

 relate

-0.64

 Mour

-0.63

 spect

-0.62

POSITIVE LOGITS

vernment

1.06

arant

1.04

ossip

1.01

errilla

1.00

gu

0.96

idelines

0.93

arantine

0.93

ipers

0.92

arding

0.91

ueless

0.90

Activations Density 0.003%

words related to disguise or hiding

the word "gu" in various contexts

No Comments

No Known Activations