INDEX

Explanations

phrases indicating future events that will not happen

oai_token-act-pair · gpt-3.5-turbo

the negative contraction "won't."

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

 newcom

-0.72

 beginnings

-0.71

anwhile

-0.69

 background

-0.64

 Background

-0.61

 populated

-0.58

 backdrop

-0.57

 compan

-0.56

 advisors

-0.56

amb

-0.56

POSITIVE LOGITS

't

2.11

kish

1.25

cest

1.17

ky

1.13

Â´

1.08

ks

1.01

now

0.99

kies

0.96

etsk

0.93

cheon

0.93

Activations Density 0.034%

phrases indicating future events that will not happen

the negative contraction "won't."

No Comments

No Known Activations