INDEX

Explanations

words related to statistical analysis and evaluation

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 specialize

-0.67

zzy

-0.62

phia

-0.59

grain

-0.57

 Myster

-0.56

 dresses

-0.55

 Kardash

-0.55

 loves

-0.54

 monog

-0.53

 queens

-0.52

POSITIVE LOGITS

etheless

0.73

''.

0.72

JECT

0.70

.(

0.68

TPS

0.67

`.

0.64

Redditor

0.63

 Orwell

0.62

emonium

0.62

 sqor

0.59

Activations Density 0.963%

words related to statistical analysis and evaluation

No Comments

No Known Activations