INDEX

Explanations

the word "or" along with a high activation value

oai_token-act-pair · gpt-3.5-turbo

conditional phrases indicating alternatives or lack thereof

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

ETS

-0.71

edu

-0.69

efer

-0.67

auga

-0.67

edia

-0.64

 Sorce

-0.63

 horm

-0.63

imes

-0.62

aylor

-0.62

KEN

-0.62

POSITIVE LOGITS

chard

1.14

Else

1.10

nam

1.08

ifice

1.03

acles

1.00

acle

0.98

chid

0.97

 else

0.93

 otherwise

0.92

ific

0.91

Activations Density 0.129%

the word "or" along with a high activation value

conditional phrases indicating alternatives or lack thereof

No Comments

No Known Activations