INDEX

Explanations

numbers written with a mix of uppercase and lowercase characters

oai_token-act-pair · gpt-3.5-turbo

references to large quantities or populations

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

 behavi

-0.75

UCT

-0.71

senal

-0.70

Oregon

-0.67

abus

-0.64

 disposition

-0.63

 Dock

-0.63

netflix

-0.62

ORN

-0.62

 arrang

-0.62

POSITIVE LOGITS

eteen

0.96

een

0.84

teen

0.79

aneous

0.78

angular

0.73

ths

0.71

 consulted

0.71

 consecutive

0.68

ofi

0.67

CFR

0.66

Activations Density 0.073%

numbers written with a mix of uppercase and lowercase characters

references to large quantities or populations

No Comments

No Known Activations