INDEX

Explanations

proper nouns, specifically news and media outlets

oai_token-act-pair · gpt-3.5-turbo

references to exclusive content or entities

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ake

-0.83

park

-0.79

ats

-0.74

ety

-0.74

iger

-0.74

 interpreter

-0.74

oline

-0.72

imer

-0.72

 clipboard

-0.70

ok

-0.69

POSITIVE LOGITS

CLUS

2.96

lishing

1.45

lance

1.28

qua

1.26

olitics

1.20

 HuffPost

1.16

 POLITICO

1.14

ÄŁ

1.12

 Dhabi

1.11

ishment

1.00

Activations Density 0.057%

proper nouns, specifically news and media outlets

references to exclusive content or entities

No Comments

No Known Activations