INDEX

Explanations

acronyms and abbreviations related to institutions or organizations

oai_token-act-pair · gpt-3.5-turbo

abbreviations or acronyms in the text

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

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 batter

-0.68

vag

-0.64

ãĤ°

-0.64

down

-0.63

lets

-0.63

***

-0.62

 gorilla

-0.62

ighed

-0.61

zzle

-0.61

tenance

-0.61

POSITIVE LOGITS

SA

1.20

ESA

1.18

RC

1.17

NH

1.16

HA

1.16

CA

1.15

AC

1.15

JC

1.15

IC

1.14

WC

1.13

Activations Density 0.152%

acronyms and abbreviations related to institutions or organizations

abbreviations or acronyms in the text

No Comments

No Known Activations