INDEX

Explanations

phrases where the word "up" is emphasized or repeated

oai_token-act-pair · gpt-3.5-turbo

instances of the word "up."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂł

-0.65

________________________

-0.65

 blinded

-0.63

 segregated

-0.61

num

-0.60

 debtor

-0.60

 Manson

-0.59

âĸ¬âĸ¬

-0.59

\\\\\\\\\\\\\\\\

-0.59

 mileage

-0.59

POSITIVE LOGITS

dates

1.43

stairs

1.26

olicy

1.17

dating

1.03

grades

1.03

graded

1.01

icult

1.00

rising

0.95

odcast

0.94

grade

0.94

Activations Density 0.027%

phrases where the word "up" is emphasized or repeated

instances of the word "up."

No Comments

No Known Activations