INDEX

Explanations

words related to locations, particularly capital cities and regions

oai_token-act-pair · gpt-3.5-turbo

references to capital cities

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

tnc

-0.86

ocket

-0.75

-->

-0.69

pite

-0.68

xon

-0.67

meier

-0.65

hiba

-0.65

agall

-0.64

henko

-0.64

atility

-0.63

POSITIVE LOGITS

 famed

0.82

 acronym

0.76

 headquartered

0.70

 birthplace

0.69

 extraord

0.68

 largest

0.68

 busiest

0.67

 precursor

0.67

 tasked

0.67

 formerly

0.67

Activations Density 0.750%

words related to locations, particularly capital cities and regions

references to capital cities

No Comments

No Known Activations