INDEX

Explanations

names of people from news sources or interviews

oai_token-act-pair · gpt-3.5-turbo

mentions of a specific person in a context related to news or interviews

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

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

 tremend

-0.95

 satell

-0.81

eatures

-0.79

 empt

-0.75

 pocket

-0.72

wcs

-0.71

 challeng

-0.71

 cumbers

-0.71

 thirst

-0.70

 stagn

-0.70

POSITIVE LOGITS

 Yeah

1.43

 Exactly

1.39

 Explain

1.35

 Well

1.28

 Alright

1.24

 Absolutely

1.20

 Okay

1.18

 Right

1.17

 Sure

1.13

 Correct

1.12

Activations Density 0.037%

names of people from news sources or interviews

mentions of a specific person in a context related to news or interviews

No Comments

No Known Activations