INDEX

Explanations

locations within a building

oai_token-act-pair · gpt-3.5-turbo

references to physical locations and structures within a space

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

CLAIM

-0.78

errors

-0.77

lessness

-0.73

gyn

-0.73

zn

-0.71

immune

-0.70

Downloadha

-0.69

Tracker

-0.69

ivia

-0.68

oglobin

-0.68

POSITIVE LOGITS

 courtyard

1.31

 plaza

1.29

 basement

1.23

 lobby

1.19

 sidewalk

1.16

 roof

1.13

 corner

1.12

 perimeter

1.10

 balcony

1.10

 annex

1.08

Activations Density 0.162%

locations within a building

references to physical locations and structures within a space

No Comments

No Known Activations