INDEX

Explanations

statements indicating surprise or unexpected information

oai_token-act-pair · gpt-3.5-turbo

phrases that convey surprising information or revelations

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 - surprise

No Comments

Negative Logits

 advoc

-0.67

 weaving

-0.66

 predicate

-0.65

 vend

-0.64

 unrestricted

-0.64

 conduit

-0.62

alach

-0.62

 guarant

-0.62

odynam

-0.62

 bluff

-0.61

POSITIVE LOGITS

 surprise

0.71

etheless

0.69

merce

0.68

shock

0.64

 Shiny

0.63

0.62

 surprises

0.62

0.61

whel

0.61

OUS

0.60

Activations Density 0.193%

statements indicating surprise or unexpected information

phrases that convey surprising information or revelations

No Comments

No Known Activations