INDEX

Explanations

terms related to technology and company products

oai_token-act-pair · gpt-3.5-turbo

structured text elements and their relationships

oai_token-act-pair · gpt-4o-mini Triggered by @bot

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Bod

-0.95

 poss

-0.80

Bol

-0.79

 heav

-0.77

Hok

-0.77

-0.75

USA

-0.74

 nodd

-0.72

 Duty

-0.72

 booklet

-0.71

POSITIVE LOGITS

ade

0.96

"]=>

0.93

Ì

0.90

imore

0.86

igan

0.85

resh

0.84

Ń·

0.82

idem

0.81

acan

0.80

riz

0.80

Activations Density 0.227%

terms related to technology and company products

structured text elements and their relationships

No Comments

No Known Activations