INDEX

Explanations

features or characteristics with high importance or impact

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ummer

-0.69

omething

-0.67

uca

-0.67

creen

-0.67

cale

-0.66

ilver

-0.65

ync

-0.64

ombie

-0.63

uman

-0.63

ecause

-0.61

POSITIVE LOGITS

pool

1.10

lists

1.08

 list

1.03

 combinations

1.03

 count

1.01

sets

0.99

 assortment

0.99

 lists

0.98

 pools

0.97

 repertoire

0.96

Activations Density 0.745%

features or characteristics with high importance or impact

No Comments

No Known Activations

features or characteristics with high importance or impact

No Comments

No Known Activations