INDEX

Explanations

legalese terms and legal actions

oai_token-act-pair · gpt-3.5-turbo

phrases related to legal frameworks and regulations

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

pretty

-0.70

everything

-0.64

laughs

-0.64

 Everywhere

-0.63

 Tackle

-0.62

 Canaver

-0.61

Everything

-0.60

esides

-0.60

anmar

-0.60

 Everything

-0.59

POSITIVE LOGITS

or

1.19

 thereto

1.13

 thereof

1.05

 Plaint

1.03

 lawfully

0.96

nor

0.96

soever

0.90

 lawful

0.89

 oneself

0.89

 persons

0.89

Activations Density 0.599%

legalese terms and legal actions

phrases related to legal frameworks and regulations

No Comments

No Known Activations