INDEX

Explanations

phrases indicating something that has never been seen or experienced before

oai_token-act-pair · gpt-3.5-turbo

phrases that denote unprecedented experiences or events

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Moral

-0.64

 Partial

-0.58

 Block

-0.56

 tense

-0.56

 alias

-0.55

 Russell

-0.55

 Maria

-0.54

 Shell

-0.54

RM

-0.54

antic

-0.53

POSITIVE LOGITS

 imaginable

0.92

cedented

0.90

 hitherto

0.88

 ever

0.85

ever

0.85

 existed

0.84

 dreamed

0.83

 anywhere

0.81

velt

0.79

thinkable

0.78

Activations Density 0.191%

phrases indicating something that has never been seen or experienced before

phrases that denote unprecedented experiences or events

No Comments

No Known Activations