INDEX

Explanations

prices and figures stated in a dollar format

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 =================================================================

-0.71

pmwiki

-0.69

================================================================

-0.65

soDeliveryDate

-0.65

 confessions

-0.64

thodox

-0.61

leaders

-0.61

 Judicial

-0.60

 Architects

-0.60

================================

-0.59

POSITIVE LOGITS

XY

0.90

foo

0.87

foo

0.83

 somebody

0.81

 ninety

0.79

['

0.75

 oranges

0.71

 blah

0.71

 tomorrow

0.71

 magically

0.71

Activations Density 0.658%

prices and figures stated in a dollar format

No Comments

No Known Activations