INDEX

Explanations

mentions of different types of rooms

oai_token-act-pair · gpt-3.5-turbo

locations within indoor spaces

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

actionDate

-0.85

ename

-0.83

thood

-0.83

ploy

-0.77

oras

-0.76

uga

-0.76

ovsky

-0.74

sen

-0.73

ulus

-0.72

İĭ

-0.70

POSITIVE LOGITS

 door

1.30

 doorway

1.24

 doors

1.18

 floor

1.15

 drawer

1.11

 window

1.08

 hallway

1.05

 entrance

1.02

 stall

1.00

 ceiling

1.00

Activations Density 0.176%

mentions of different types of rooms

locations within indoor spaces

No Comments

No Known Activations