INDEX

Explanations

character names or titles followed by a number

oai_token-act-pair · gpt-3.5-turbo

references to animated characters or content related to animation

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

 goodwill

-0.67

Vaj

-0.65

 favorably

-0.64

 constrained

-0.63

 minded

-0.62

 disapp

-0.59

 Mechdragon

-0.59

 downhill

-0.59

 conditional

-0.59

kit

-0.58

POSITIVE LOGITS

ampions

1.47

ampion

1.47

ocolate

1.43

aos

1.34

orus

1.34

osen

1.33

annels

1.28

unks

1.28

apters

1.26

amber

1.25

Activations Density 0.028%

character names or titles followed by a number

references to animated characters or content related to animation

No Comments

No Known Activations