INDEX

Explanations

phrases related to marking or classifying information

oai_token-act-pair · gpt-3.5-turbo

instances of things being labeled or categorized

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

imates

-0.88

isance

-0.81

yre

-0.77

urus

-0.74

isner

-0.73

 breathed

-0.73

rouch

-0.70

ADS

-0.70

ersive

-0.68

rouse

-0.67

POSITIVE LOGITS

furt

0.81

 signs

0.79

Sta

0.76

 marked

0.75

ocument

0.75

 milestones

0.71

marked

0.67

 markings

0.67

mast

0.64

 Pupp

0.63

Activations Density 0.025%

phrases related to marking or classifying information

instances of things being labeled or categorized

No Comments

No Known Activations