INDEX

Explanations

explicitly offensive and controversial language or concepts

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

 satell

-0.95

 staggered

-0.88

 glim

-0.85

 challeng

-0.85

 acquisitions

-0.78

everal

-0.78

 bulky

-0.78

 accumulated

-0.77

athered

-0.75

 planned

-0.75

POSITIVE LOGITS

 Doesn

1.19

 Saying

1.18

 Seriously

1.16

 Secondly

1.15

 Nope

1.13

 Shame

1.12

 Explain

1.10

 Does

1.10

Isn

1.09

 Otherwise

1.08

Activations Density 0.428%

explicitly offensive and controversial language or concepts

No Comments

No Known Activations