INDEX

Explanations

instances of the word "written."

oai_token-act-pair · gpt-3.5-turbo

instances of the word "written" and its variations in the text

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

auder

-0.73

lymp

-0.72

aband

-0.71

 disrupt

-0.67

alore

-0.62

 Expand

-0.62

oras

-0.62

tackle

-0.62

vantage

-0.61

buster

-0.61

POSITIVE LOGITS

 written

3.76

 Written

2.57

written

2.53

Written

2.13

 rewritten

2.13

 writing

1.87

 penned

1.72

 write

1.69

 printed

1.64

writing

1.57

Activations Density 0.014%

instances of the word "written."

instances of the word "written" and its variations in the text

No Comments

No Known Activations