INDEX

Explanations

publication-related phrases such as titles, author names, and dates

oai_token-act-pair · gpt-3.5-turbo

academic references and citations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 cooled

-0.72

 sprung

-0.71

 rented

-0.71

 raided

-0.70

 thankfully

-0.70

 departed

-0.69

 wandered

-0.69

 cheaply

-0.68

 stricken

-0.68

 luckily

-0.67

POSITIVE LOGITS

Understanding

1.46

Introduction

1.45

 Understanding

1.30

Effects

1.26

 Introduction

1.26

 Principles

1.26

Compar

1.22

 Comparison

1.22

 Effects

1.19

 Towards

1.18

Activations Density 0.238%

publication-related phrases such as titles, author names, and dates

academic references and citations

No Comments

No Known Activations