INDEX

Explanations

phrases indicating desire or preference

oai_token-act-pair · gpt-3.5-turbo

expressions of desire or personal wants

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Archer

-0.60

 âĹı

-0.54

++++++++++++++++

-0.54

 FANTASY

-0.53

 Indust

-0.53

 âĨ

-0.53

 Transparency

-0.53

Geh

-0.51

Orn

-0.48

 Hard

-0.48

POSITIVE LOGITS

illet

0.59

ocument

0.59

strap

0.58

 destined

0.58

liest

0.56

 accompl

0.52

 deem

0.52

 muster

0.51

 wrought

0.51

 lawfully

0.50

Activations Density 1.431%

phrases indicating desire or preference

expressions of desire or personal wants

No Comments

No Known Activations