INDEX

Explanations

terminology related to protection or being safeguarded

oai_token-act-pair · gpt-3.5-turbo

references to the concept of protection or safeguarding

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

izzle

-0.64

itch

-0.64

rum

-0.62

iem

-0.62

 palate

-0.61

ISTORY

-0.61

iga

-0.61

ology

-0.61

gram

-0.61

alities

-0.60

POSITIVE LOGITS

 protected

3.79

protected

2.91

 unprotected

2.25

 shielded

1.95

 protection

1.73

 protects

1.68

 protecting

1.65

 protect

1.65

 guarded

1.61

 safegu

1.60

Activations Density 0.021%

terminology related to protection or being safeguarded

references to the concept of protection or safeguarding

No Comments

No Known Activations