INDEX

Explanations

references to the word "Duke"

oai_token-act-pair · gpt-3.5-turbo

references to the term "Duke."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

hedon

-0.71

ãģĻ

-0.67

iped

-0.67

cells

-0.67

ulations

-0.66

iations

-0.65

ional

-0.64

served

-0.62

agency

-0.62

ters

-0.61

POSITIVE LOGITS

hyde

1.08

halla

0.80

ongyang

0.75

onna

0.74

 Duke

0.73

 Bride

0.71

ham

0.71

lac

0.71

hole

0.70

isine

0.67

Activations Density 0.021%

references to the word "Duke"

references to the term "Duke."

No Comments

No Known Activations