INDEX

Explanations

phrases related to news events or people

oai_token-act-pair · gpt-3.5-turbo

various plural forms or suffixes of words that indicate groups or collections

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 exception

-0.74

 reinforcement

-0.65

ASED

-0.64

JECT

-0.63

âĢ¢âĢ¢âĢ¢âĢ¢

-0.63

ctors

-0.62

cyt

-0.60

opathic

-0.60

Assembly

-0.59

 benz

-0.59

POSITIVE LOGITS

peed

1.43

ilver

1.36

hirt

1.33

paces

1.26

pace

1.25

aurus

1.25

omething

1.24

mith

1.21

hare

1.18

hots

1.18

Activations Density 0.169%

phrases related to news events or people

various plural forms or suffixes of words that indicate groups or collections

No Comments

No Known Activations