INDEX

Explanations

proper nouns related to politics and public figures

oai_token-act-pair · gpt-3.5-turbo

references to political figures and contexts

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 yearly

-0.58

 anew

-0.57

 annually

-0.56

 forever

-0.55

 perpetually

-0.54

 ALWAYS

-0.53

 every

-0.53

ennial

-0.52

 routinely

-0.52

 constantly

-0.51

POSITIVE LOGITS

autions

0.49

 defence

0.46

 humour

0.44

 colour

0.43

 flavour

0.43

avour

0.42

ãģ¾

0.42

 tyre

0.42

rounder

0.42

 Sense

0.40

Activations Density 1.267%

proper nouns related to politics and public figures

references to political figures and contexts

No Comments

No Known Activations