INDEX

Explanations

high-level abstract concepts or terms

oai_token-act-pair · gpt-3.5-turbo

concepts and keywords related to abstraction in technology and programming

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

rique

-0.83

otype

-0.67

 zoning

-0.66

itted

-0.64

bred

-0.64

 excerpts

-0.63

iless

-0.62

essee

-0.62

lace

-0.61

otic

-0.61

POSITIVE LOGITS

mson

1.05

geist

0.81

 Huss

0.81

hemer

0.76

 Kant

0.75

framework

0.75

$$$$

0.75

âĢķ

0.74

knife

0.73

ubuntu

0.73

Activations Density 0.029%

high-level abstract concepts or terms

concepts and keywords related to abstraction in technology and programming

No Comments

No Known Activations