INDEX

Explanations

numerical quantities followed by a unit of measurement

oai_token-act-pair · gpt-3.5-turbo

instances of the word "runs" in various contexts

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

ether

-0.68

itive

-0.67

illus

-0.64

ase

-0.64

athering

-0.63

Crime

-0.63

atti

-0.62

iquette

-0.62

OME

-0.60

Letter

-0.59

POSITIVE LOGITS

 runs

3.53

 Runs

2.88

runs

2.53

run

2.22

ran

1.98

Run

1.80

run

1.70

Run

1.64

 running

1.53

RUN

1.51

Activations Density 0.014%

numerical quantities followed by a unit of measurement

instances of the word "runs" in various contexts

No Comments

No Known Activations