INDEX

Explanations

ordinal numbers indicating ranking or positions

oai_token-act-pair · gpt-3.5-turbo

references to rankings and numerical lists

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

uala

-0.85

elcome

-0.76

joy

-0.69

acet

-0.69

tap

-0.68

yrim

-0.68

ajor

-0.66

ueller

-0.66

ington

-0.66

igham

-0.65

POSITIVE LOGITS

 earners

1.02

%:

0.92

 rankings

0.90

 contenders

0.89

%,

0.85

 performers

0.83

 ranking

0.81

%.

0.81

 percent

0.81

0.79

Activations Density 0.051%

ordinal numbers indicating ranking or positions

references to rankings and numerical lists

No Comments

No Known Activations