INDEX

Explanations

strong verbs indicating caution or action recommendations

oai_token-act-pair · gpt-3.5-turbo

phrases that involve warnings, demands, and calls for action or change

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

ecast

-0.84

ocument

-0.82

imum

-0.77

ixt

-0.76

erenn

-0.75

elta

-0.75

imester

-0.72

efeated

-0.71

tyard

-0.70

iral

-0.69

POSITIVE LOGITS

 accordingly

1.17

 consequently

0.83

 encour

0.78

 regrets

0.76

 apolog

0.75

 caution

0.74

 accompanying

0.72

 vowed

0.71

 apologise

0.71

 apologize

0.71

Activations Density 0.470%

strong verbs indicating caution or action recommendations

phrases that involve warnings, demands, and calls for action or change

No Comments

No Known Activations