INDEX

Explanations

phrases indicating that more information is available on a particular topic

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate the act of learning or acquiring information

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

albeit

-0.83

cffffcc

-0.78

ãĥĨãĤ£

-0.76

typically

-0.75

 alone

-0.71

erity

-0.71

mouth

-0.69

ynes

-0.69

HF

-0.68

usually

-0.68

POSITIVE LOGITS

how

1.28

why

1.22

WHY

0.97

How

0.87

why

0.86

HOW

0.85

 what

0.85

 whats

0.80

 selecting

0.80

 this

0.79

Activations Density 0.160%

phrases indicating that more information is available on a particular topic

phrases that indicate the act of learning or acquiring information

No Comments

No Known Activations