INDEX

Explanations

proper nouns related to individuals, particularly those named Francis

oai_token-act-pair · gpt-3.5-turbo

mentions of the name "Francis."

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ruciating

-0.69

mble

-0.68

phis

-0.67

nesota

-0.67

ocaust

-0.66

restling

-0.64

iable

-0.63

 Wrestling

-0.61

 Wrestle

-0.61

Õ

-0.60

POSITIVE LOGITS

 Francis

1.12

Franc

1.06

 Xavier

0.84

olini

0.84

 Benedict

0.74

Ã©r

0.73

 Bacon

0.73

icer

0.69

olate

0.69

 pope

0.68

Activations Density 0.010%

proper nouns related to individuals, particularly those named Francis

mentions of the name "Francis."

No Comments

No Known Activations