INDEX

Explanations

expressions of gratitude or condolences on behalf of a group

oai_token-act-pair · gpt-3.5-turbo

expressions of gratitude or acknowledgment in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 monop

-0.77

uggle

-0.73

 stuff

-0.65

 infiltration

-0.65

iddle

-0.64

 trickle

-0.64

 hybrids

-0.64

IQ

-0.64

 mole

-0.64

optional

-0.63

POSITIVE LOGITS

 THANK

1.37

 thank

1.32

 congratulations

1.29

 apologies

1.22

 Thank

1.16

 rejoice

1.16

Thank

1.16

 congratulate

1.13

 commend

1.13

Congratulations

1.13

Activations Density 1.277%

expressions of gratitude or condolences on behalf of a group

expressions of gratitude or acknowledgment in various contexts

No Comments

No Known Activations