INDEX

Explanations

parentheses followed by numbers

oai_token-act-pair · gpt-3.5-turbo

references to lists or enumerations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

iera

-0.74

imore

-0.71

SPONSORED

-0.67

 haun

-0.65

vale

-0.65

 Vaugh

-0.64

angan

-0.63

/*

-0.63

 guard

-0.61

 wreck

-0.60

POSITIVE LOGITS

 Firstly

1.05

 Introduction

1.01

 Whilst

0.94

 Approximately

0.93

 Lack

0.92

 Beginning

0.92

 Unless

0.91

 Whether

0.91

 Immediately

0.90

 Assuming

0.90

Activations Density 0.092%

parentheses followed by numbers

references to lists or enumerations

No Comments

No Known Activations