INDEX

Explanations

terms related to magical elements or actions

oai_token-act-pair · gpt-3.5-turbo

references to magic

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

ETF

-0.82

 Bund

-0.75

ourced

-0.73

lihood

-0.73

APD

-0.73

QB

-0.72

eters

-0.71

pread

-0.71

Uk

-0.71

 Nguyen

-0.70

POSITIVE LOGITS

 magic

3.63

magic

2.86

 magical

2.28

Magic

2.25

 Magic

2.16

 magician

1.94

 sorcery

1.93

 wizards

1.64

 miracles

1.62

 wizard

1.61

Activations Density 0.011%

terms related to magical elements or actions

references to magic

No Comments

No Known Activations