INDEX

Explanations

instances of the word "Gir" with varying levels of activations, suggesting a potential focus on this specific term

oai_token-act-pair · gpt-3.5-turbo

mentions of "Gir" and variations of that term, indicative of a specific entity or organization

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

ittee

-0.81

ãĥĩãĤ£

-0.73

ãĥ¼ãĥĨ

-0.70

 Contra

-0.68

opsy

-0.68

 Archdemon

-0.64

eal

-0.63

 CLSID

-0.63

edia

-0.63

 Dragonbound

-0.62

POSITIVE LOGITS

Gir

1.16

affe

1.03

aldi

0.85

 Pengu

0.83

oux

0.81

kees

0.76

arde

0.74

aiman

0.74

vana

0.74

imar

0.73

Activations Density 0.009%

instances of the word "Gir" with varying levels of activations, suggesting a potential focus on this specific term

mentions of "Gir" and variations of that term, indicative of a specific entity or organization

No Comments

No Known Activations