INDEX

Explanations

instances of the word "windows"

oai_token-act-pair · gpt-3.5-turbo

references to "windows" 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 @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

avez

-0.80

ghan

-0.72

IGH

-0.72

ARA

-0.71

oku

-0.67

zin

-0.66

zing

-0.64

REDACTED

-0.64

yss

-0.63

 recomm

-0.62

POSITIVE LOGITS

 windows

1.02

wip

1.01

windows

0.98

 pane

0.96

 window

0.85

 sill

0.85

glass

0.84

bars

0.83

door

0.82

 openings

0.81

Activations Density 0.010%

instances of the word "windows"

references to "windows" in various contexts

No Comments

No Known Activations