INDEX

Explanations

words describing the location or position of objects or people

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "sits."

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

 attribution

-0.63

 dangerous

-0.62

 fatal

-0.61

 serious

-0.60

 Method

-0.60

 humane

-0.56

 Limit

-0.56

 mutual

-0.56

clusively

-0.56

 exploit

-0.56

POSITIVE LOGITS

 sits

3.36

sit

1.80

 hangs

1.65

sat

1.60

 resides

1.58

 occupies

1.56

 leans

1.56

 lays

1.47

 awaits

1.45

 stands

1.44

Activations Density 0.010%

words describing the location or position of objects or people

occurrences of the word "sits."

No Comments

No Known Activations