INDEX

Explanations

commands or instructions directed at the reader

oai_token-act-pair · gpt-3.5-turbo

phrases addressing the reader directly regarding experiences or knowledge

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

stown

-0.67

antle

-0.66

conn

-0.63

 Flavoring

-0.62

æ©

-0.62

worth

-0.61

margin

-0.61

borg

-0.60

cial

-0.60

Apps

-0.59

POSITIVE LOGITS

 wanna

0.73

ocument

0.73

ILLE

0.72

raints

0.70

 accidentally

0.69

 choke

0.67

NPR

0.67

 curious

0.66

 recess

0.65

 handy

0.65

Activations Density 0.049%

commands or instructions directed at the reader

phrases addressing the reader directly regarding experiences or knowledge

No Comments

No Known Activations