INDEX

Explanations

words related to qualities such as elegance, courage, and justice

oai_token-act-pair · gpt-3.5-turbo

cohesive conjunctions that suggest unity or connection within a context

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

-0.89

RN

-0.82

çīĪ

-0.81

ricanes

-0.80

 Nare

-0.80

rolet

-0.78

emort

-0.78

sbm

-0.75

acent

-0.74

odcast

-0.73

POSITIVE LOGITS

 elegance

1.45

 humility

1.37

 creativity

1.36

 heroism

1.34

 bravery

1.30

 innocence

1.29

 righteousness

1.27

 generosity

1.24

 openness

1.24

 rationality

1.23

Activations Density 0.411%

words related to qualities such as elegance, courage, and justice

cohesive conjunctions that suggest unity or connection within a context

No Comments

No Known Activations