INDEX

Explanations

phrases related to exploring or investigating options or situations

oai_token-act-pair · gpt-3.5-turbo

various forms of the word "explain."

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

 Cups

-0.68

 Helsinki

-0.67

 Mens

-0.67

rier

-0.67

li

-0.66

 ledger

-0.66

 cholesterol

-0.64

lam

-0.64

 Berk

-0.63

 Dane

-0.63

POSITIVE LOGITS

expl

3.67

Expl

3.17

 Expl

2.07

 unexpl

1.72

 expl

1.46

impl

1.42

 Exploration

1.29

 decomp

1.28

 Explosive

1.24

 Explos

1.18

Activations Density 0.009%

phrases related to exploring or investigating options or situations

various forms of the word "explain."

No Comments

No Known Activations