INDEX

Explanations

the word "loyal" with a high activation

oai_token-act-pair · gpt-3.5-turbo

the concept of loyalty

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

EVA

-0.74

OUT

-0.71

 Virus

-0.67

 Autism

-0.66

 Drugs

-0.65

 Schwarz

-0.65

 Genetics

-0.65

FER

-0.64

 Clinic

-0.64

LOD

-0.64

POSITIVE LOGITS

 loyal

1.06

itiz

1.05

 allegiance

0.95

 loyalty

0.93

alty

0.91

ties

0.91

atile

0.88

 adherent

0.88

iciary

0.87

 servant

0.85

Activations Density 0.007%

the word "loyal" with a high activation

the concept of loyalty

No Comments

No Known Activations