INDEX

Explanations

mentions of abbreviations in the format "Int" followed by a number, potentially related to intelligence or interaction

oai_token-act-pair · gpt-3.5-turbo

instances of the term "int."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

-0.75

hyde

-0.73

è¦ļéĨĴ

-0.72

wordpress

-0.72

geon

-0.71

borg

-0.71

mith

-0.70

SHIP

-0.69

illard

-0.68

ĸļ

-0.68

POSITIVE LOGITS

ensity

1.18

elligence

1.15

ellect

1.14

ention

1.10

elligent

1.03

ENTION

1.02

ellig

0.96

ended

0.94

ensive

0.89

estinal

0.87

Activations Density 0.012%

mentions of abbreviations in the format "Int" followed by a number, potentially related to intelligence or interaction

instances of the term "int."

No Comments

No Known Activations