INDEX

Explanations

phrases related to violent or alarming actions

oai_token-act-pair · gpt-3.5-turbo

phrases related to threats and harm

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

uge

-0.75

 âĢº

-0.66

zbollah

-0.65

oglu

-0.63

 pioneered

-0.62

particularly

-0.62

 largeDownload

-0.60

oiler

-0.60

Palestinian

-0.58

very

-0.58

POSITIVE LOGITS

his

1.01

 hers

1.01

 himself

0.96

 theirs

0.95

him

0.93

 herself

0.86

her

0.79

him

0.72

 them

0.72

the

0.72

Activations Density 0.899%

phrases related to violent or alarming actions

phrases related to threats and harm

No Comments

No Known Activations