INDEX

Explanations

proper nouns related to news organizations

oai_token-act-pair · gpt-3.5-turbo

references to news publications, particularly "The Washington Post"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

sbm

-0.83

sed

-0.76

gone

-0.74

stood

-0.68

 directions

-0.63

wana

-0.62

hra

-0.62

milo

-0.61

pex

-0.61

plot

-0.60

POSITIVE LOGITS

 Staff

0.78

 Wire

0.78

pei

0.71

Buy

0.70

 Â·

0.69

 Toledo

0.68

ebus

0.68

 toggle

0.65

ipeg

0.65

Associated

0.65

Activations Density 0.061%

proper nouns related to news organizations

references to news publications, particularly "The Washington Post"

No Comments

No Known Activations