INDEX

Explanations

words related to deception or false information

oai_token-act-pair · gpt-3.5-turbo

mentions of hoaxes and pranks

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

bourg

-0.68

 Borders

-0.68

aws

-0.66

ailable

-0.65

uv

-0.65

uner

-0.62

udeau

-0.62

bilt

-0.62

oyal

-0.61

asper

-0.60

POSITIVE LOGITS

 hoax

1.03

sters

0.87

²¾

0.84

erella

0.81

ually

0.80

ishly

0.79

edly

0.79

es

0.78

ulence

0.75

ed

0.75

Activations Density 0.029%

words related to deception or false information

mentions of hoaxes and pranks

No Comments

No Known Activations