INDEX

Explanations

phrases or words related to packaging or grouping items together

oai_token-act-pair · gpt-3.5-turbo

references to multipacks of products

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 notoriously

-0.63

 fateful

-0.63

izabeth

-0.61

 recourse

-0.60

 plagued

-0.59

 egreg

-0.59

 disastrous

-0.59

hor

-0.58

 irre

-0.57

 troublesome

-0.57

POSITIVE LOGITS

pack

1.44

Pack

1.15

rats

0.98

packs

0.89

eers

0.84

owitz

0.83

 Pack

0.82

ascript

0.82

Parser

0.80

enger

0.79

Activations Density 0.007%

phrases or words related to packaging or grouping items together

references to multipacks of products

No Comments

No Known Activations