INDEX

Explanations

apologies or regrets expressed either explicitly or implied in sentences

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 Opposition

-0.82

 Quarterly

-0.77

 Slate

-0.76

 Quartz

-0.72

 Argon

-0.72

 Berger

-0.71

 Reconstruction

-0.71

 Securities

-0.70

 Mobil

-0.70

 Lyons

-0.69

POSITIVE LOGITS

 dont

1.38

lol

1.31

 didnt

1.24

sorry

1.22

yeah

1.21

please

1.20

thanks

1.19

thank

1.19

 haha

1.17

 doesnt

1.16

Activations Density 0.364%

apologies or regrets expressed either explicitly or implied in sentences

No Comments

No Known Activations