INDEX

Explanations

capitalized terms that denote exclusive content or special features

oai_token-act-pair · gpt-3.5-turbo

instances of a specific prefix or category related to exclusivity

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

¶ħ

-0.78

kowski

-0.76

Sut

-0.75

kson

-0.71

erson

-0.67

enance

-0.66

¿½

-0.66

kamp

-0.65

ãĤ¡

-0.64

 livelihood

-0.63

POSITIVE LOGITS

PRESS

1.31

TERN

1.30

CLUS

1.27

PORT

1.24

CEPT

1.23

PER

1.13

POS

1.08

cellent

1.01

TER

1.00

ACT

1.00

Activations Density 0.012%

capitalized terms that denote exclusive content or special features

instances of a specific prefix or category related to exclusivity

No Comments

No Known Activations