INDEX

Explanations

objects or entities that can be identified, collected, traded, or donated

oai_token-act-pair · gpt-3.5-turbo

references to items 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

iversity

-0.77

 rhythms

-0.69

stretched

-0.68

 valleys

-0.68

govern

-0.68

adia

-0.65

alus

-0.65

 jails

-0.64

cia

-0.64

 Valhalla

-0.63

POSITIVE LOGITS

 item

3.63

item

2.39

 Item

2.33

 items

2.31

Item

2.16

 Items

1.99

items

1.77

Items

1.73

 artifact

1.37

 object

1.32

Activations Density 0.015%

objects or entities that can be identified, collected, traded, or donated

references to items in various contexts

No Comments

No Known Activations