INDEX

Explanations

abbreviations or acronyms starting with "SP" followed by a number

oai_token-act-pair · gpt-3.5-turbo

instances of the abbreviation "SP" followed by various numbers

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

âĶģ

-0.73

hower

-0.73

ablishment

-0.73

sbm

-0.73

algia

-0.72

gerald

-0.71

Ã§a

-0.71

FactoryReloaded

-0.70

ãĥĥãĥĪ

-0.68

emis

-0.67

POSITIVE LOGITS

EED

1.16

ONSORED

1.09

OIL

1.02

ARK

0.96

IR

0.93

SP

0.91

ORT

0.87

LC

0.86

anish

0.85

ECT

0.84

Activations Density 0.005%

abbreviations or acronyms starting with "SP" followed by a number

instances of the abbreviation "SP" followed by various numbers

No Comments

No Known Activations