INDEX

Explanations

phrases related to instructions or announcements, especially related to events or organizations

oai_token-act-pair · gpt-3.5-turbo

terms related to applications or formal procedures

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

jack

-0.75

comb

-0.65

rock

-0.65

Hor

-0.64

 Junior

-0.63

far

-0.63

meet

-0.61

 grade

-0.60

 confid

-0.60

win

-0.59

POSITIVE LOGITS

ATION

3.95

ATIONS

3.38

ATED

2.54

ATOR

2.28

ATING

2.22

ations

2.15

ATIONAL

2.11

ATIVE

2.10

ation

1.94

ATES

1.93

Activations Density 0.010%

phrases related to instructions or announcements, especially related to events or organizations

terms related to applications or formal procedures

No Comments

No Known Activations