INDEX

Explanations

phrases related to technological messages or instructions

oai_token-act-pair · gpt-3.5-turbo

certain punctuation or formatting marks

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

Negative Logits

terness

-0.89

thal

-0.72

vertis

-0.70

meal

-0.67

Tea

-0.64

sein

-0.64

ennial

-0.63

 frivolous

-0.62

agogue

-0.62

ractions

-0.62

POSITIVE LOGITS

ira

0.65

INK

0.64

"+

0.62

lected

0.61

 Gonz

0.61

worn

0.59

 Armored

0.59

Ctrl

0.59

behind

0.58

 wielded

0.58

Activations Density 0.000%

phrases related to technological messages or instructions

certain punctuation or formatting marks

No Known Activations