INDEX

Explanations

words related to physical structures or locations, specifically rows and columns

oai_token-act-pair · gpt-3.5-turbo

references to rows and columns, particularly in a structural or organizational context

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 outing

-0.87

 whistle

-0.77

emark

-0.74

 mark

-0.73

venture

-0.71

 deal

-0.71

alg

-0.71

emaker

-0.70

eg

-0.68

afe

-0.67

POSITIVE LOGITS

 columns

2.75

 rows

2.75

 tiers

1.53

 partitions

1.52

 floors

1.50

 piles

1.43

 fields

1.42

 entrances

1.40

 formations

1.40

 stacks

1.39

Activations Density 0.043%

words related to physical structures or locations, specifically rows and columns

references to rows and columns, particularly in a structural or organizational context

No Comments

No Known Activations