INDEX

Explanations

parts related to structured text, such as document sections and titles

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

berus

-0.69

anca

-0.67

apons

-0.60

gently

-0.59

asting

-0.57

 speakers

-0.57

acha

-0.57

 cumbers

-0.57

 gaping

-0.57

okia

-0.57

POSITIVE LOGITS

ners

1.25

icularly

1.22

icular

1.20

nered

1.18

ridge

1.18

isans

1.13

ially

1.10

icles

1.07

icipated

1.03

ner

1.03

Activations Density 0.023%

parts related to structured text, such as document sections and titles

No Comments

No Known Activations