INDEX

Explanations

the word "partly" followed by a high score activation

oai_token-act-pair · gpt-3.5-turbo

the phrase "partly" and related contexts that indicate partial causation or responsibility

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

(@

-0.68

mac

-0.64

eria

-0.63

idget

-0.61

reetings

-0.59

 Attack

-0.59

-0.58

FFFF

-0.58

ipes

-0.58

!!"

-0.57

POSITIVE LOGITS

 partly

3.34

 partially

2.62

 largely

1.66

 chiefly

1.59

 principally

1.53

 mainly

1.52

 mostly

1.42

mostly

1.37

 loosely

1.31

 primarily

1.25

Activations Density 0.011%

the word "partly" followed by a high score activation

the phrase "partly" and related contexts that indicate partial causation or responsibility

No Comments

No Known Activations