INDEX

Explanations

phrases related to news headlines

oai_token-act-pair · gpt-3.5-turbo

uppercase letter sequences that suggest names or places

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

 Diaz

-0.65

 Franks

-0.65

iors

-0.63

selves

-0.61

kson

-0.61

fe

-0.59

 Germans

-0.57

stood

-0.56

self

-0.55

 bushes

-0.55

POSITIVE LOGITS

MAN

1.22

AN

1.14

ANS

1.14

LAND

1.13

COL

1.13

EN

1.13

CLAIM

1.12

VIEW

1.12

VER

1.10

VILLE

1.10

Activations Density 0.075%

phrases related to news headlines

uppercase letter sequences that suggest names or places

No Comments

No Known Activations