INDEX

Explanations

phrases indicating calls to action or suggestions for specific actions

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "call" in various forms, indicating demands or requests in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

riks

-0.62

 ILCS

-0.61

kus

-0.59

ãĥĹ

-0.57

mental

-0.56

ãĥ¼ãĥĨ

-0.56

å¥³

-0.56

Ö¼

-0.55

é¾įå

-0.55

âĢ¢âĢ¢âĢ¢âĢ¢

-0.55

POSITIVE LOGITS

 attention

1.20

for

1.00

 forth

0.92

 upon

0.92

ously

0.91

enance

0.87

phas

0.81

 into

0.81

 Attention

0.80

 bullshit

0.75

Activations Density 0.054%

phrases indicating calls to action or suggestions for specific actions

occurrences of the word "call" in various forms, indicating demands or requests in the text

No Comments

No Known Activations