INDEX

Explanations

mentions of permission or authorization

oai_token-act-pair · gpt-3.5-turbo

phrases indicating permission or capability

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

 gloom

-0.71

 headline

-0.65

 crying

-0.63

 terrible

-0.61

 gloomy

-0.59

 sweat

-0.58

 bloody

-0.58

 grave

-0.58

sob

-0.57

 disappointment

-0.56

POSITIVE LOGITS

 allows

3.07

 enables

2.40

 Allows

2.17

 lets

2.12

 allow

2.06

Allows

2.05

 facilitates

2.01

 ensures

1.96

allows

1.94

 provides

1.91

Activations Density 0.021%

mentions of permission or authorization

phrases indicating permission or capability

No Comments

No Known Activations