INDEX

Explanations

phrases related to deception and trickery

oai_token-act-pair · gpt-3.5-turbo

various forms of the word "deceive."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

Yor

-0.71

 Brands

-0.67

 Bots

-0.64

ingen

-0.63

 Targ

-0.62

 Aires

-0.62

 Polo

-0.62

 Zamb

-0.60

 Atkinson

-0.60

CHA

-0.59

POSITIVE LOGITS

ffect

1.04

ither

0.95

ptive

0.93

pt

0.92

emonic

0.92

iving

0.91

ased

0.90

fficient

0.88

ivable

0.87

astrous

0.86

Activations Density 0.094%

phrases related to deception and trickery

various forms of the word "deceive."

No Comments

No Known Activations