INDEX

Explanations

the word "whole" in phrases

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

osponsors

-0.90

acers

-0.72

none

-0.70

ummies

-0.69

orthy

-0.69

chairs

-0.68

KER

-0.68

acons

-0.66

ANS

-0.65

cs

-0.64

POSITIVE LOGITS

 thing

1.36

heartedly

1.31

hearted

1.21

 ordeal

1.19

 affair

1.08

 damn

0.98

 idea

0.97

 endeavor

0.91

 debacle

0.90

thing

0.90

Activations Density 0.033%

the word "whole" in phrases

No Comments

No Known Activations