INDEX

Explanations

names, concepts, or phrases that the reader may not have heard of before

oai_token-act-pair · gpt-3.5-turbo

phrases indicating unfamiliarity or lack of awareness about something

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

uckles

-0.70

Crystal

-0.70

estyles

-0.64

 teasp

-0.64

 Notting

-0.61

 blender

-0.60

pour

-0.59

 selfie

-0.59

essional

-0.58

ĸļ

-0.58

POSITIVE LOGITS

/+

0.88

0.76

DragonMagazine

0.71

ãĥīãĥ©ãĤ´ãĥ³

0.70

/-

0.69

ten

0.65

æĿ

0.64

FB

0.64

Ds

0.63

tx

0.63

Activations Density 0.079%

names, concepts, or phrases that the reader may not have heard of before

phrases indicating unfamiliarity or lack of awareness about something

No Comments

No Known Activations