INDEX

Explanations

titles, ranks, positions, and roles referring to people or entities

oai_token-act-pair · gpt-3.5-turbo

phrases that refer to titles and positions of authority

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

umerable

-0.83

venants

-0.79

iland

-0.79

jad

-0.75

raints

-0.73

forms

-0.71

lements

-0.71

onga

-0.71

ples

-0.69

itions

-0.69

POSITIVE LOGITS

 medi

0.76

 holder

0.74

 champion

0.73

¥µ

0.71

 outsider

0.70

 champ

0.69

 cellar

0.68

 savior

0.68

 martyr

0.68

assador

0.66

Activations Density 0.123%

titles, ranks, positions, and roles referring to people or entities

phrases that refer to titles and positions of authority

No Comments

No Known Activations