INDEX

Explanations

phrases indicating causality or conditionality

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

yna

-0.69

elve

-0.66

CVE

-0.65

 pione

-0.63

Virgin

-0.63

inois

-0.63

uty

-0.63

isively

-0.62

atl

-0.61

atri

-0.61

POSITIVE LOGITS

 they

1.33

 THEY

1.10

 something

1.08

 someone

1.06

 there

1.04

 somebody

1.02

you

1.01

 everything

0.98

it

0.94

 theirs

0.91

Activations Density 0.235%

phrases indicating causality or conditionality

No Comments

No Known Activations