INDEX

Explanations

references to things described as being unwanted

oai_token-act-pair · gpt-3.5-turbo

references to unwanted or unwelcome experiences and situations

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

boards

-0.72

 Strategy

-0.65

istry

-0.64

ept

-0.61

 theory

-0.61

Science

-0.61

omics

-0.60

 Meridian

-0.59

 Commentary

-0.59

 motto

-0.58

POSITIVE LOGITS

 unwanted

3.82

 unwelcome

2.46

 undesirable

2.12

 undes

1.93

 unintended

1.78

 unnecessary

1.44

 inappropriate

1.37

 pesky

1.33

 invasive

1.25

 unprotected

1.24

Activations Density 0.012%

references to things described as being unwanted

references to unwanted or unwelcome experiences and situations

No Comments

No Known Activations