INDEX

Explanations

website URLs or text starting with the letters "wr"

oai_token-act-pair · gpt-3.5-turbo

references to the abbreviation "WR" and related terms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 happ

-0.74

Afee

-0.70

 Transparency

-0.68

 Viet

-0.67

 SHOW

-0.63

 YEAR

-0.62

 Authorities

-0.61

PAC

-0.61

 Turing

-0.60

TOM

-0.60

POSITIVE LOGITS

wr

1.44

angler

1.05

inker

0.95

ights

0.93

ath

0.92

heed

0.92

wright

0.85

adoes

0.84

abbit

0.83

uth

0.83

Activations Density 0.013%

website URLs or text starting with the letters "wr"

references to the abbreviation "WR" and related terms

No Comments

No Known Activations