INDEX

Explanations

numbers mentioned within a text, especially relating to quantities or statistics

oai_token-act-pair · gpt-3.5-turbo

references to numerical data or statistics

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

aan

-0.71

SPONSORED

-0.71

reddit

-0.67

linger

-0.64

reath

-0.62

atom

-0.61

aine

-0.60

esson

-0.60

ired

-0.58

ifa

-0.58

POSITIVE LOGITS

 numbers

3.81

 Numbers

3.04

Numbers

2.56

 percentages

2.02

 number

1.99

 totals

1.72

number

1.70

 figures

1.70

Number

1.68

 numbering

1.66

Activations Density 0.014%

numbers mentioned within a text, especially relating to quantities or statistics

references to numerical data or statistics

No Comments

No Known Activations