INDEX

Explanations

phrases related to formal or technical specifications

oai_token-act-pair · gpt-3.5-turbo

phrases related to the experience of interruptions or distractions

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

GU

-0.64

ZIP

-0.64

 Gmail

-0.63

MAP

-0.63

 pyramid

-0.62

BBC

-0.62

 Laurent

-0.62

 fodder

-0.62

 Samar

-0.62

 apple

-0.61

POSITIVE LOGITS

1.19

ski

1.08

sic

1.04

sat

1.04

sent

0.99

ï¸ı

0.98

never

0.96

leans

0.95

felt

0.93

ses

0.92

Activations Density 0.200%

phrases related to formal or technical specifications

phrases related to the experience of interruptions or distractions

No Comments

No Known Activations