INDEX

Explanations

words related to trends or renditions, potentially in various contexts like finance, artistry, or storytelling

oai_token-act-pair · gpt-3.5-turbo

instances of the word "rend" and its variations, indicating a focus on destruction or separation

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

OTOS

-0.68

 Kaufman

-0.64

ãĥīãĥ©

-0.63

 Warsaw

-0.61

 Lovecraft

-0.59

keley

-0.58

GY

-0.58

aiman

-0.58

avez

-0.58

cules

-0.55

POSITIVE LOGITS

ezvous

1.56

rend

1.46

erer

1.16

eering

0.94

inen

0.94

erers

0.92

lyn

0.92

enance

0.90

icator

0.86

ened

0.85

Activations Density 0.009%

words related to trends or renditions, potentially in various contexts like finance, artistry, or storytelling

instances of the word "rend" and its variations, indicating a focus on destruction or separation

No Comments

No Known Activations