INDEX

Explanations

words related to items or entities not further specified, or that are part of lists or categories of objects

oai_token-act-pair · gpt-3.5-turbo

references to additional items or concepts in a list-like context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

each

-0.69

frog

-0.66

utenant

-0.64

 shampoo

-0.64

ruary

-0.63

 Lithuania

-0.63

bike

-0.63

wark

-0.63

culosis

-0.62

uay

-0.61

POSITIVE LOGITS

worldly

1.43

 assorted

1.35

 forms

1.05

 kinds

0.99

 similar

0.97

 types

0.95

 aspects

0.91

 pertinent

0.90

 sorts

0.89

 goodies

0.88

Activations Density 0.071%

words related to items or entities not further specified, or that are part of lists or categories of objects

references to additional items or concepts in a list-like context

No Comments

No Known Activations