INDEX

Explanations

mentions of the word "Wat" with varying degree of specificity

oai_token-act-pair · gpt-3.5-turbo

references to a specific team or entity named "Watford."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 prevailing

-0.75

 Haas

-0.70

pmwiki

-0.69

 grape

-0.65

 appropriation

-0.65

interstitial

-0.64

 blunt

-0.63

 reconciliation

-0.63

 accounting

-0.62

wcs

-0.62

POSITIVE LOGITS

terson

1.42

Wat

1.10

chers

1.02

chens

1.01

combe

1.00

pering

0.98

cher

0.98

chet

0.96

ches

0.96

atche

0.95

Activations Density 0.009%

mentions of the word **"Wat"** with varying degree of specificity

references to a specific team or entity named "Watford."

No Comments

No Known Activations

mentions of the word "Wat" with varying degree of specificity