INDEX

Explanations

mentions of the word "ub" with varying activation levels

oai_token-act-pair · gpt-3.5-turbo

references to "ub" as a recurrent pattern or theme

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

 Lauder

-0.78

 ORIG

-0.72

 drift

-0.67

 Atlantic

-0.66

alez

-0.65

 Irma

-0.63

FUL

-0.62

 Burnett

-0.61

backer

-0.61

agher

-0.61

POSITIVE LOGITS

lishing

1.26

bing

1.21

rious

1.15

lisher

1.12

lique

1.11

bed

1.10

bles

1.10

bish

1.10

lish

1.05

ilant

1.04

Activations Density 0.034%

mentions of the word "ub" with varying activation levels

references to "ub" as a recurrent pattern or theme

No Comments

No Known Activations