INDEX

Explanations

highly abstract or conceptual phrases related to ideas or theories

oai_token-act-pair · gpt-3.5-turbo

words or phrases indicating significant change or disruption

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 juggling

-0.71

adm

-0.65

 Rhod

-0.62

 Hemisphere

-0.59

SO

-0.59

 Monroe

-0.57

 blond

-0.57

 puff

-0.56

 lighter

-0.56

 homebrew

-0.56

POSITIVE LOGITS

should

1.31

were

1.22

was

1.19

would

1.19

must

1.19

has

1.16

could

1.15

may

1.14

might

1.12

comes

1.09

Activations Density 0.253%

highly abstract or conceptual phrases related to ideas or theories

words or phrases indicating significant change or disruption

No Comments

No Known Activations