INDEX

Explanations

the continuation markers "Continued" and "Below" in news articles

oai_token-act-pair · gpt-3.5-turbo

segment titles or headings often used to introduce additional content in articles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

onomic

-0.62

Wan

-0.62

wives

-0.60

merce

-0.59

usterity

-0.58

asters

-0.58

collar

-0.58

morrow

-0.56

ergic

-0.56

tackle

-0.55

POSITIVE LOGITS

 Continued

0.72

 Below

0.71

 Transcript

0.71

 McCain

0.67

ï

0.62

 VIDEOS

0.59

 CNBC

0.59

 srfAttach

0.59

 Morning

0.58

 Flake

0.58

Activations Density 0.013%

the continuation markers "Continued" and "Below" in news articles

segment titles or headings often used to introduce additional content in articles

No Comments

No Known Activations