INDEX

Explanations

references to academic publishing activities like "published", "reviewed", and "online"

oai_token-act-pair · gpt-3.5-turbo

terms related to the publication and review of academic or online content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 whist

-0.76

 whistle

-0.65

Geh

-0.63

 Magnet

-0.63

 Manifest

-0.63

 strings

-0.62

 Shake

-0.60

 Module

-0.60

 Infrastructure

-0.60

 caps

-0.59

POSITIVE LOGITS

published

2.51

reviewed

2.14

online

1.96

atoon

1.37

eden

1.19

review

1.18

Reviewed

1.15

released

1.14

pole

0.97

ingen

0.95

Activations Density 0.066%

references to academic publishing activities like "published", "reviewed", and "online"

terms related to the publication and review of academic or online content

No Comments

No Known Activations