INDEX

Explanations

details about things that people like or enjoy

oai_token-act-pair · gpt-3.5-turbo

instances of the word "likes."

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

Â§

-0.68

 borne

-0.64

examination

-0.64

ANCE

-0.63

imony

-0.61

mit

-0.61

athon

-0.59

Impl

-0.58

sequence

-0.58

 AMERICA

-0.58

POSITIVE LOGITS

 likes

3.91

 Likes

1.82

 loves

1.62

 liked

1.56

 prefers

1.49

 hates

1.38

 liking

1.37

 wants

1.27

 favourites

1.20

 enjoys

1.12

Activations Density 0.012%

details about things that people like or enjoy

instances of the word "likes."

No Comments

No Known Activations