INDEX

Explanations

action verbs indicating things to be done or needed

oai_token-act-pair · gpt-3.5-turbo

phrases indicating requirements or expectations for readers

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

minecraft

-0.73

 Adin

-0.69

 Goodbye

-0.67

Kal

-0.66

Tok

-0.62

00000000

-0.61

ocular

-0.61

 Fundamental

-0.61

legal

-0.60

 Pegasus

-0.60

POSITIVE LOGITS

 notice

0.96

 yourself

0.83

 noticed

0.80

 hear

0.79

merce

0.79

 need

0.74

 wonder

0.73

 yourselves

0.73

 realize

0.72

 encount

0.72

Activations Density 0.152%

action verbs indicating things to be done or needed

phrases indicating requirements or expectations for readers

No Comments

No Known Activations