INDEX

Explanations

proper nouns or acronyms being activated are not representative enough to identify a specific pattern or theme

oai_token-act-pair · gpt-3.5-turbo

instances of the abbreviation "SE" 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 @ 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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

uron

-0.63

 anthrop

-0.63

bob

-0.62

 cannons

-0.62

 hero

-0.61

 upright

-0.61

 Inferno

-0.60

ibn

-0.60

 wire

-0.59

 prop

-0.59

POSITIVE LOGITS

SE

4.33

SE

1.86

SEA

1.77

se

1.74

NE

1.51

SI

1.48

SEE

1.44

SA

1.43

ME

1.39

CE

1.37

Activations Density 0.010%

proper nouns or acronyms being activated are not representative enough to identify a specific pattern or theme

instances of the abbreviation "SE" in various contexts

No Comments

No Known Activations