INDEX

Explanations

phrases related to providing insight or a brief preview of future events

oai_token-act-pair · gpt-3.5-turbo

references to glimpses or previews of information or concepts

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

Not in Any Lists

No Comments

Negative Logits

hement

-0.75

ients

-0.71

KK

-0.70

cial

-0.69

ubs

-0.67

die

-0.66

lees

-0.66

arently

-0.64

eches

-0.64

 depended

-0.63

POSITIVE LOGITS

 whats

0.91

 sorts

0.83

helm

0.77

 theirs

0.73

 ours

0.73

 doom

0.71

 what

0.69

 physiology

0.69

 reality

0.68

how

0.66

Activations Density 0.187%

phrases related to providing insight or a brief preview of future events

references to glimpses or previews of information or concepts

No Comments

No Known Activations