INDEX

Explanations

phrases related to principles, beliefs, or teachings

oai_token-act-pair · gpt-3.5-turbo

references to religious or philosophical doctrines

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

netflix

-0.81

hots

-0.77

RGB

-0.74

umber

-0.69

Bron

-0.67

angan

-0.62

SB

-0.62

ASH

-0.62

 Shiny

-0.61

umers

-0.60

POSITIVE LOGITS

 doctrine

3.69

 Doctrine

2.89

 doctrines

2.88

 dogma

2.02

 doctr

1.98

 theology

1.71

 orthodoxy

1.63

 teachings

1.52

 theory

1.42

 heresy

1.35

Activations Density 0.016%

phrases related to principles, beliefs, or teachings

references to religious or philosophical doctrines

No Comments

No Known Activations