INDEX

Explanations

words or phrases related to clicking on links or navigating websites

oai_token-act-pair · gpt-3.5-turbo

phrases related to online actions and interactions, particularly clicking and navigating to websites

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

unker

-0.67

 improv

-0.66

uster

-0.63

 morale

-0.63

dam

-0.63

oustic

-0.62

 mids

-0.62

 fracturing

-0.61

 dams

-0.60

 deepening

-0.60

POSITIVE LOGITS

pages

1.05

 menus

1.05

 listings

1.01

 scroll

0.97

 Browse

0.96

 menu

0.93

0.87

 accessed

0.85

 homepage

0.85

menu

0.85

Activations Density 0.532%

words or phrases related to clicking on links or navigating websites

phrases related to online actions and interactions, particularly clicking and navigating to websites

No Comments

No Known Activations