INDEX

Explanations

words related to holding beliefs or positions

oai_token-act-pair · gpt-3.5-turbo

expressions of possessing or maintaining views and positions of power

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

————

-0.78

issance

-0.71

lease

-0.71

lez

-0.68

ese

-0.67

ombies

-0.66

ghan

-0.66

endix

-0.65

FTWARE

-0.64

ibel

-0.63

POSITIVE LOGITS

 sway

1.27

 onto

1.11

 accountable

0.99

 steady

0.92

 hostage

0.90

 captive

0.87

 dear

0.83

overs

0.81

hold

0.81

 secrets

0.81

Activations Density 0.040%

words related to holding beliefs or positions

expressions of possessing or maintaining views and positions of power

No Comments

No Known Activations