INDEX

Explanations

the word "ub" with varying positive activation values

oai_token-act-pair · gpt-3.5-turbo

instances of the substring "ub" 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 @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 Lauder

-0.79

 ozone

-0.66

 Irma

-0.65

 Noir

-0.65

backer

-0.63

 drift

-0.62

alez

-0.62

 grounding

-0.62

cig

-0.62

 ORIG

-0.62

POSITIVE LOGITS

lishing

1.37

bing

1.27

lisher

1.20

bish

1.18

lish

1.16

lique

1.12

rious

1.10

bed

1.10

untu

1.05

bies

1.05

Activations Density 0.019%

the word "ub" with varying positive activation values

instances of the substring "ub" in various contexts

No Comments

No Known Activations