INDEX

Explanations

mentions of the word "walnut"

oai_token-act-pair · gpt-3.5-turbo

references to walnuts and conservative viewpoints

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Meter

-0.78

 Dante

-0.75

 Static

-0.73

Cao

-0.72

Ion

-0.71

¿½

-0.71

 Boyle

-0.69

 Magnet

-0.69

 Donovan

-0.68

 Hung

-0.68

POSITIVE LOGITS

wal

3.43

Wal

1.24

âĢİ

1.20

mah

1.16

ervative

1.15

fox

1.10

wal

1.04

 almonds

1.03

ay

1.03

 pist

0.98

Activations Density 0.041%

mentions of the word "walnut"

references to walnuts and conservative viewpoints

No Comments

No Known Activations