INDEX

Explanations

dates represented as the month "January" followed by a day and potentially other words or numbers

oai_token-act-pair · gpt-3.5-turbo

references to the date "January 9."

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Reviewer

-0.73

 diaper

-0.68

ropolitan

-0.66

 Bearing

-0.66

 manual

-0.65

ACTED

-0.64

puff

-0.62

SLI

-0.61

cipline

-0.61

 diapers

-0.60

POSITIVE LOGITS

uary

1.11

itor

1.05

itors

1.03

vier

1.02

eway

0.87

esville

0.87

iors

0.87

ice

0.86

uine

0.83

owitz

0.83

Activations Density 0.006%

dates represented as the month "January" followed by a day and potentially other words or numbers

references to the date "January 9."

No Comments

No Known Activations