INDEX

Explanations

contractions of words to identify emphasis, assertion, or intent in text

oai_token-act-pair · gpt-3.5-turbo

instances of the word "re."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

Iv

-0.68

 dotted

-0.63

 levels

-0.62

 Bulg

-0.58

 Skies

-0.58

 chalk

-0.58

 faces

-0.57

ATL

-0.57

 anywhere

-0.57

 hinges

-0.56

POSITIVE LOGITS

re

3.93

res

2.12

reb

2.04

RE

1.84

rea

1.71

rely

1.68

rer

1.64

ret

1.58

rep

1.57

rev

1.55

Activations Density 0.056%

contractions of words to identify emphasis, assertion, or intent in text

instances of the word "re."

No Comments

No Known Activations