INDEX

Explanations

mentions of evolutionary themes or concepts

oai_token-act-pair · gpt-3.5-turbo

references to the word "Evil" and its variants

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

 Gord

-0.68

 prescribed

-0.67

 Sara

-0.66

 Sharif

-0.66

 ends

-0.65

 ribbon

-0.65

gru

-0.65

 Luxem

-0.64

 brass

-0.64

XL

-0.64

POSITIVE LOGITS

Ev

3.88

EV

1.53

Iv

1.38

Ir

1.36

Ec

1.36

Dar

1.28

ev

1.25

Ay

1.23

Av

1.21

Factor

1.12

Activations Density 0.017%

mentions of evolutionary themes or concepts

references to the word "Evil" and its variants

No Comments

No Known Activations