INDEX

Explanations

proper names or titles with the prefix "Er"

oai_token-act-pair · gpt-3.5-turbo

mentions of individuals and their roles or identities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

yip

-0.81

uesday

-0.74

jri

-0.74

ongyang

-0.73

okemon

-0.73

ccoli

-0.72

 Flavoring

-0.71

iless

-0.71

benefit

-0.70

isSpecialOrderable

-0.69

POSITIVE LOGITS

ipel

0.88

0.72

×Ļ

0.72

 Petersen

0.70

×¨

0.68

0.67

 Kramer

0.64

 omission

0.63

uary

0.62

tainment

0.62

Activations Density 0.101%

proper names or titles with the prefix "Er"

mentions of individuals and their roles or identities

No Comments

No Known Activations