INDEX

Explanations

the word "Orient" or variations of it

oai_token-act-pair · gpt-3.5-turbo

terms related to orientation and spatial positioning

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

ukemia

-0.85

CLOSE

-0.79

claw

-0.76

 Mellon

-0.75

SourceFile

-0.74

 plaque

-0.69

eva

-0.69

Gw

-0.68

Sco

-0.68

 repeal

-0.68

POSITIVE LOGITS

 Orient

3.60

orient

2.74

 orient

2.41

 Oriental

1.87

rient

1.63

ori

1.37

 Blanc

1.30

Ori

1.24

rients

1.23

Hoy

1.19

Activations Density 0.040%

the word "Orient" or variations of it

terms related to orientation and spatial positioning

No Comments

No Known Activations