INDEX

Explanations

locations within a building

oai_token-act-pair · gpt-3.5-turbo

locations and structures within a scene

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

thood

-0.76

arus

-0.74

actionDate

-0.64

nesia

-0.62

oneliness

-0.61

edIn

-0.61

Southern

-0.59

planet

-0.58

usterity

-0.58

vasive

-0.57

POSITIVE LOGITS

 doorway

1.16

 doors

1.16

 ceiling

1.15

 entrance

1.10

 aisle

1.06

 railing

1.04

 door

1.04

 balcony

1.03

 hallway

1.03

 ledge

0.99

Activations Density 0.194%

locations within a building

locations and structures within a scene

No Comments

No Known Activations