INDEX

Explanations

phrases marked by the word "wit."

oai_token-act-pair · gpt-3.5-turbo

instances of the word "wit" and its variants 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 @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 obligated

-0.72

ĪĴ

-0.71

 enrollment

-0.66

Ü

-0.65

ĺħ

-0.62

cknow

-0.62

 lobbying

-0.62

 Shuttle

-0.61

 predatory

-0.60

 educating

-0.59

POSITIVE LOGITS

hered

1.19

chery

1.13

glers

0.94

wit

0.94

bol

0.89

nesses

0.86

ener

0.85

ivities

0.83

nder

0.83

ilt

0.82

Activations Density 0.009%

phrases marked by the word "wit."

instances of the word "wit" and its variants in various contexts

No Comments

No Known Activations