INDEX

Explanations

words related to actions leading to a change or transformation, which include the concept of broadening or expanding

oai_token-act-pair · gpt-3.5-turbo

variations of the suffix "ened" 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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 schools

-0.60

 Banks

-0.59

 Towers

-0.57

 finding

-0.55

 captcha

-0.54

 collecting

-0.54

 confirm

-0.54

 Million

-0.54

LAT

-0.51

 school

-0.51

POSITIVE LOGITS

ened

4.58

ening

2.95

eners

2.20

ener

1.95

ens

1.95

enment

1.72

ENS

1.40

ered

1.34

EN

1.27

ended

1.26

Activations Density 0.008%

words related to actions leading to a change or transformation, which include the concept of broadening or expanding

variations of the suffix "ened" in various contexts

No Comments

No Known Activations