INDEX

Explanations

mentions of the word "Electron" or related terms

oai_token-act-pair · gpt-3.5-turbo

references to elections and electronic voting

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

Lens

-0.77

ACK

-0.69

OAD

-0.67

 Flavoring

-0.67

acca

-0.67

BuyableInstoreAndOnline

-0.67

 Airways

-0.66

ãĤ¬

-0.64

Kit

-0.63

========

-0.62

POSITIVE LOGITS

rified

1.15

rification

1.15

roc

1.13

ron

1.13

rons

1.12

rics

1.06

rum

1.04

rated

1.03

rical

1.02

oral

0.99

Activations Density 0.016%

mentions of the word "Electron" or related terms

references to elections and electronic voting

No Comments

No Known Activations