INDEX

Explanations

mentions of love or affection towards someone

oai_token-act-pair · gpt-3.5-turbo

instances of the word "loved" and expressions of affection or emotional connections

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

•layer 0 - love

No Comments

Negative Logits

statement

-0.69

======

-0.64

 deposition

-0.64

report

-0.63

dm

-0.61

admin

-0.61

DM

-0.61

system

-0.59

 blocking

-0.59

minimum

-0.58

POSITIVE LOGITS

 loved

3.89

 loves

2.06

 hated

1.85

 liked

1.79

 love

1.72

 beloved

1.63

 cherished

1.60

 loving

1.60

 disliked

1.48

 admired

1.46

Activations Density 0.018%

mentions of love or affection towards someone

instances of the word "loved" and expressions of affection or emotional connections

No Comments

No Known Activations