INDEX

Explanations

references to legal and justice-related terms, such as "Law & Order" and "law and order"

oai_token-act-pair · gpt-3.5-turbo

conjunctions and phrases that emphasize connections or associations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

auga

-0.78

unts

-0.73

 Cosponsors

-0.73

visual

-0.71

sites

-0.69

mith

-0.66

href

-0.64

gest

-0.64

nces

-0.62

ryce

-0.62

POSITIVE LOGITS

 order

1.03

order

1.02

 ORDER

0.83

 Order

0.78

 Disorder

0.76

ordered

0.75

 disorder

0.70

 reconciliation

0.67

 mercy

0.66

 regulation

0.65

Activations Density 0.114%

references to legal and justice-related terms, such as "Law & Order" and "law and order"

conjunctions and phrases that emphasize connections or associations

No Comments

No Known Activations