INDEX

Explanations

phrases related to revealing or uncovering something hidden

oai_token-act-pair · gpt-3.5-turbo

repeated references to curtains and veils, indicating themes of concealment or barriers

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 nomine

-0.72

etically

-0.70

nesota

-0.69

*/(

-0.68

ILY

-0.68

OTAL

-0.68

ebin

-0.67

ciating

-0.67

livious

-0.66

ishops

-0.65

POSITIVE LOGITS

 curtain

1.26

 curtains

1.22

glass

1.00

 veil

0.80

bridge

0.76

 Veil

0.75

ãĤ¦ãĤ¹

0.72

AGE

0.72

rod

0.71

acle

0.71

Activations Density 0.009%

phrases related to revealing or uncovering something hidden

repeated references to curtains and veils, indicating themes of concealment or barriers

No Comments

No Known Activations