INDEX

Explanations

words related to hidden information or confidential matters

oai_token-act-pair · gpt-3.5-turbo

phrases and words related to secrets

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

orer

-0.67

MLA

-0.64

urch

-0.64

ariat

-0.63

phe

-0.62

 Huss

-0.62

CBO

-0.62

unicip

-0.61

Paw

-0.61

 Nationwide

-0.60

POSITIVE LOGITS

 secrets

4.11

 Secrets

2.57

 secret

2.06

 mysteries

1.96

 secrecy

1.65

secret

1.63

Secret

1.55

 treasures

1.52

 truths

1.50

 surprises

1.40

Activations Density 0.017%

words related to hidden information or confidential matters

phrases and words related to secrets

No Comments

No Known Activations