INDEX

Explanations

specific sequences of characters that typically represent the end of a text, possibly indicating a new section or a break in the document

oai_token-act-pair · gpt-3.5-turbo

significant temporal markers or dates

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 amnesty

-0.68

 tradem

-0.63

undai

-0.60

onite

-0.58

 intervened

-0.58

 sacrificed

-0.56

 eleph

-0.55

 referen

-0.55

 acquies

-0.55

illions

-0.54

POSITIVE LOGITS

CLOSE

0.96

Still

0.90

WASHINGTON

0.89

About

0.86

Quote

0.85

Overview

0.85

BACK

0.84

Details

0.83

Media

0.83

Posted

0.83

Activations Density 0.064%

specific sequences of characters that typically represent the end of a text, possibly indicating a new section or a break in the document

significant temporal markers or dates

No Comments

No Known Activations