INDEX

Explanations

word sequences with "IC" followed by a number

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

wards

-0.73

wich

-0.71

itism

-0.71

ments

-0.62

geons

-0.62

selves

-0.62

 Shutterstock

-0.62

 miracles

-0.61

hip

-0.61

spread

-0.60

POSITIVE LOGITS

ANN

1.36

BM

1.18

IJ

0.99

ICLE

0.98

Ns

0.91

ritical

0.90

ONY

0.89

ategory

0.87

LE

0.87

trl

0.86

Activations Density 0.030%

word sequences with "IC" followed by a number

No Comments

No Known Activations