INDEX

Explanations

phrases related to actions or statements that stand out distinctly

oai_token-act-pair · gpt-3.5-turbo

the term 'Out' and related phrases indicating a release or exit

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 precedent

-0.62

 stem

-0.60

 religiously

-0.59

du

-0.59

 canon

-0.59

fir

-0.58

 torn

-0.58

 consult

-0.58

 glac

-0.57

 dear

-0.57

POSITIVE LOGITS

Out

3.45

OUT

2.00

Out

1.95

out

1.91

outs

1.67

OUT

1.63

Outside

1.48

Off

1.41

Over

1.29

Down

1.27

Activations Density 0.010%

phrases related to actions or statements that stand out distinctly

the term 'Out' and related phrases indicating a release or exit

No Comments

No Known Activations