INDEX

Explanations

mentions of royalty or leadership

oai_token-act-pair · gpt-3.5-turbo

references to royalty, specifically kings and queens

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ATIONAL

-0.71

chell

-0.69

ename

-0.67

 Disability

-0.66

mble

-0.66

ogene

-0.65

inguished

-0.64

 Squirrel

-0.64

closed

-0.63

imilar

-0.63

POSITIVE LOGITS

 king

1.25

esses

1.10

pin

1.06

pins

1.01

naire

0.97

doms

0.96

lord

0.94

fish

0.92

 kings

0.89

uin

0.87

Activations Density 0.007%

mentions of royalty or leadership

references to royalty, specifically kings and queens

No Comments

No Known Activations