INDEX

Explanations

words related to familial relationships, specifically "Brother"

oai_token-act-pair · gpt-3.5-turbo

references to "brother" and "sister" 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 @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

UTC

-0.90

=#

-0.79

aukee

-0.69

 Helsinki

-0.68

udeau

-0.65

CBC

-0.65

rawdownloadcloneembedreportprint

-0.64

PLIED

-0.64

doi

-0.64

 convey

-0.63

POSITIVE LOGITS

 Brother

1.50

Brother

1.16

thren

0.99

 Brothers

0.98

 Sister

0.95

brother

0.86

 Nath

0.82

 Sisters

0.77

ishop

0.76

axter

0.76

Activations Density 0.007%

words related to familial relationships, specifically "Brother"

references to "brother" and "sister" in various contexts

No Comments

No Known Activations