INDEX

Explanations

locations or institutions named "Fort" followed by a single-word identifier

oai_token-act-pair · gpt-3.5-turbo

references to specific military locations or installations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

Magikarp

-0.71

lihood

-0.67

 subscribing

-0.66

Companies

-0.65

FREE

-0.63

soever

-0.62

Accessory

-0.61

âĶģ

-0.60

Publisher

-0.59

UGH

-0.59

POSITIVE LOGITS

uitous

1.38

itude

1.27

ieth

1.10

dinand

1.08

 Fort

1.04

resses

1.02

 Lauderdale

1.02

oise

0.97

itud

0.95

isec

0.88

Activations Density 0.013%

locations or institutions named "Fort" followed by a single-word identifier

references to specific military locations or installations

No Comments

No Known Activations