INDEX

Explanations

words related to unique identifiers or titles, such as names, initials, or acronyms

oai_token-act-pair · gpt-3.5-turbo

the letter "P" at the beginning of words

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

adm

-0.75

 Bots

-0.68

roy

-0.67

 Deter

-0.63

 phyl

-0.62

 variables

-0.61

 diplom

-0.61

 Miko

-0.61

flows

-0.60

 pleas

-0.60

POSITIVE LOGITS

ossible

1.40

redict

1.37

ossibly

1.35

odcast

1.32

rotein

1.31

ERSON

1.31

ractical

1.30

ermanent

1.28

owered

1.26

ilot

1.25

Activations Density 0.032%

words related to unique identifiers or titles, such as names, initials, or acronyms

the letter "P" at the beginning of words

No Comments

No Known Activations