INDEX

Explanations

Lisp code snippets

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Configuration

neuronpedia/gpt2-small__res_scl-ajt/6-res_scl-ajt

Prompts (Dashboard)

12,288 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

46,080

Data Type

torch.float32

Hook Point

blocks.6.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

apollo-research/Skylion007-openwebtext-tokenizer-gpt2

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

•Layer 6 UMAP region: Mostly-local line at bottom - local

No Comments

Negative Logits

 Transcript

-0.31

 spoilers

-0.30

 Lauder

-0.30

usterity

-0.29

 Poverty

-0.28

 Perspect

-0.28

 Emails

-0.28

 Murdoch

-0.27

 Labour

-0.27

 Electoral

-0.26

POSITIVE LOGITS

Accessory

0.31

ensis

0.30

âĹ¼

0.29

pecially

0.29

 outper

0.27

ppel

0.27

liner

0.27

onite

0.27

ãĤ¤ãĥĪ

0.27

âĵĺ

0.26

Activations Density 13.108%

Lisp code snippets

No Comments

No Known Activations