INDEX

Explanations

terms related to being dominant or in a position of power

oai_token-act-pair · gpt-3.5-turbo

terms related to "defendants" and "plates" 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 @ 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

 Must

-0.71

mul

-0.69

 Normal

-0.68

 Plays

-0.65

Len

-0.64

 Shooting

-0.64

 Witches

-0.64

 Kills

-0.63

 Loss

-0.63

 Zombies

-0.63

POSITIVE LOGITS

endant

3.17

endants

2.75

plate

2.13

plates

1.79

ande

1.38

hend

1.35

antle

1.16

 bracelet

1.11

allion

1.10

pter

1.09

Activations Density 0.039%

terms related to being dominant or in a position of power

terms related to "defendants" and "plates" in various contexts

No Comments

No Known Activations