INDEX

Explanations

references to executing commands or creating users/groups in a software setting

oai_token-act-pair · gpt-3.5-turbo

the word "and" 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 @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

meet

-0.86

deen

-0.76

Gaza

-0.72

jured

-0.71

conduct

-0.71

CNN

-0.71

ilitary

-0.70

hur

-0.68

å¥³

-0.67

Shut

-0.67

POSITIVE LOGITS

 optionally

1.16

 hence

1.11

 consequently

1.08

 therefore

1.06

 thus

1.05

 thereby

1.02

 then

1.01

rogen

1.00

rogens

1.00

romeda

0.97

Activations Density 0.789%

references to executing commands or creating users/groups in a software setting

the word "and" in various contexts

No Comments

No Known Activations