INDEX

Explanations

crime-related actions and legal terms

oai_token-act-pair · gpt-3.5-turbo

instances of the word "and" in a criminal context

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

DX

-0.79

 Powered

-0.72

isible

-0.68

Unity

-0.68

Own

-0.68

fuck

-0.68

meric

-0.67

lux

-0.67

tale

-0.66

Age

-0.65

POSITIVE LOGITS

 consequently

1.10

 therefore

1.04

 thus

1.02

 hence

1.00

 then

0.94

 possibly

0.94

 thereby

0.94

 subsequently

0.90

 urged

0.84

 presumably

0.83

Activations Density 0.622%

crime-related actions and legal terms

instances of the word "and" in a criminal context

No Comments

No Known Activations