INDEX

Explanations

phrases related to bringing something to an end or resolution

oai_token-act-pair · gpt-3.5-turbo

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

thouse

-0.68

 galleries

-0.66

DragonMagazine

-0.63

 fielded

-0.60

 coefficients

-0.60

Rect

-0.60

 nicer

-0.59

azines

-0.59

 rooft

-0.58

 finish

-0.58

POSITIVE LOGITS

angering

1.28

owment

1.22

eared

0.96

ocrine

0.84

angered

0.83

ocrin

0.83

 prematurely

0.81

angers

0.81

ogeneity

0.81

urable

0.80

Activations Density 0.044%

phrases related to bringing something to an end or resolution

No Comments

No Known Activations