INDEX

Explanations

references to different scientific journals

oai_token-act-pair · gpt-3.5-turbo

references to academic journals

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

evil

-0.73

rises

-0.72

ardless

-0.65

BuyableInstoreAndOnline

-0.65

 ours

-0.64

merce

-0.64

cream

-0.63

mph

-0.63

places

-0.62

christ

-0.62

POSITIVE LOGITS

 Journal

1.15

Journal

1.13

ournals

1.13

 Journals

0.98

Paper

0.97

 Editors

0.94

 Papers

0.93

 journal

0.92

 Review

0.86

 Entry

0.86

Activations Density 0.006%

references to different scientific journals

references to academic journals

No Known Activations