INDEX

Explanations

the word "leave."

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "LE" and its variants, likely indicating a theme of leadership or organization

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

Torrent

-0.82

Connell

-0.81

Redditor

-0.80

cephal

-0.79

Desk

-0.78

ginx

-0.74

TPPStreamerBot

-0.73

Downloadha

-0.72

flix

-0.72

zig

-0.72

POSITIVE LOGITS

LE

2.25

ME

2.19

SE

2.17

FE

2.16

CE

2.14

NO

2.11

NE

1.98

DE

1.78

TE

1.77

RE

1.76

Activations Density 0.031%

the word "leave."

occurrences of the word "LE" and its variants, likely indicating a theme of leadership or organization

No Comments

No Known Activations