INDEX

Explanations

mentions of the word "Geek" or similar variations

oai_token-act-pair · gpt-3.5-turbo

references to the name "Ge" and variations thereof

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 antibiotic

-0.70

 cottage

-0.67

 chant

-0.65

 limp

-0.64

 waged

-0.64

 checkpoints

-0.64

 causation

-0.64

 laundry

-0.63

PAT

-0.63

 dividing

-0.63

POSITIVE LOGITS

Ge

3.56

Geo

2.20

Ge

1.33

ge

1.30

 Geek

1.15

Crypt

1.07

ge

1.05

Rock

0.97

GH

0.97

geo

0.95

Activations Density 0.031%

mentions of the word "Geek" or similar variations

references to the name "Ge" and variations thereof

No Comments

No Known Activations