INDEX

Explanations

names and dates in social media posts

oai_token-act-pair · gpt-3.5-turbo

references to Twitter handles or user mentions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 6-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.6.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.6.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

 harms

-0.75

houses

-0.72

itory

-0.71

 attractions

-0.68

 pockets

-0.68

 clothes

-0.68

 criminals

-0.66

fruit

-0.64

 medicines

-0.63

Ĥª

-0.63

POSITIVE LOGITS

TPS

1.00

0.83

Twe

0.78

>]

0.71

 Ùħ

0.70

 VIDEOS

0.69

Official

0.69

VERTIS

0.68

Patch

0.67

 Originally

0.66

Activations Density 0.043%

names and dates in social media posts

references to Twitter handles or user mentions

No Comments

No Known Activations