INDEX

Explanations

references to something cherished or deeply loved

oai_token-act-pair · gpt-3.5-turbo

references to the concept of being beloved or cherished

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

•layer 5 - love

No Comments

Negative Logits

agnetic

-0.71

ulhu

-0.71

IDER

-0.71

annis

-0.70

conf

-0.69

degree

-0.69

amation

-0.68

akedown

-0.67

otomy

-0.67

uid

-0.66

POSITIVE LOGITS

 dearly

0.80

 beloved

0.76

 pets

0.74

 cherished

0.74

 Creatures

0.73

 uncond

0.69

 darling

0.68

 Ashes

0.67

 memory

0.67

 memories

0.67

Activations Density 0.014%

references to something cherished or deeply loved

references to the concept of being beloved or cherished

No Comments

No Known Activations