INDEX

Explanations

phrases related to locations, specifically the United States

oai_token-act-pair · gpt-3.5-turbo

references to the United States

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

 penet

-0.70

accompan

-0.68

operator

-0.63

inances

-0.62

lighting

-0.62

ticket

-0.61

erella

-0.61

Dog

-0.60

warts

-0.59

ioned

-0.59

POSITIVE LOGITS

 Indies

0.86

 Territories

0.81

 Isles

0.77

 Alps

0.75

 illegally

0.75

 States

0.74

 Hudson

0.70

vale

0.70

 Wasteland

0.70

 countryside

0.70

Activations Density 0.130%

phrases related to locations, specifically the United States

references to the United States

No Comments

No Known Activations