INDEX

Explanations

phrases related to urging or advising others to do or not do something

oai_token-act-pair · gpt-3.5-turbo

negative imperatives or expressions indicating prohibition

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

catentry

-0.67

Judge

-0.66

Fourth

-0.65

rouse

-0.60

Analysis

-0.60

utherford

-0.60

ItemThumbnailImage

-0.59

Higher

-0.58

ullah

-0.58

METHOD

-0.57

POSITIVE LOGITS

ï¸ı

0.73

 kidding

0.72

=""

0.69

 âĢº

0.64

 ðŁĳ

0.64

(@

0.63

ymes

0.63

ĸļ

0.62

 ðŁ

0.60

 FANTASY

0.60

Activations Density 0.645%

phrases related to urging or advising others to do or not do something

negative imperatives or expressions indicating prohibition

No Comments

No Known Activations