INDEX

Explanations

action-related terms, potentially related to standing up or standing out

oai_token-act-pair · gpt-3.5-turbo

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

-0.69

 constitu

-0.66

 Rica

-0.64

 Rican

-0.62

ļé

-0.59

 Doodle

-0.58

--------------------------------------------------------

-0.57

ROR

-0.57

EStream

-0.57

 Dusk

-0.56

POSITIVE LOGITS

alone

1.43

bys

1.11

up

1.06

off

1.01

offs

1.01

through

1.00

out

0.97

around

0.97

away

0.96

down

0.96

Activations Density 0.019%

action-related terms, potentially related to standing up or standing out

No Comments

No Known Activations