INDEX

Explanations

power-related words and phrases, especially those related to authority or control

oai_token-act-pair · gpt-3.5-turbo

references to various forms of authority and control

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Mell

-0.66

 Murd

-0.65

 Dwell

-0.64

rio

-0.63

 Mour

-0.63

rian

-0.62

riad

-0.61

 straw

-0.60

Gil

-0.59

 Cure

-0.59

POSITIVE LOGITS

 powers

1.00

hell

0.96

 conferred

0.89

Reviewer

0.87

 ashore

0.82

 delegated

0.82

 exerc

0.77

uits

0.77

 vested

0.75

 Powers

0.72

Activations Density 0.025%

power-related words and phrases, especially those related to authority or control

references to various forms of authority and control

No Comments

No Known Activations