INDEX

Explanations

locations where tragic events have occurred

oai_token-act-pair · gpt-3.5-turbo

sentences that indicate events or facts

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 tremend

-0.75

 monopol

-0.74

orah

-0.73

iron

-0.73

 worthwhile

-0.72

 challeng

-0.71

 corrid

-0.68

rael

-0.68

 infring

-0.68

 differentiation

-0.67

POSITIVE LOGITS

 Doctors

1.45

Doctors

1.42

His

1.38

His

1.34

 Authorities

1.29

 Investigators

1.26

 Prosecutors

1.18

TMZ

1.16

He

1.14

 Officials

1.12

Activations Density 0.364%

locations where tragic events have occurred

sentences that indicate events or facts

No Comments

No Known Activations