INDEX

Explanations

phrases that introduce a new topic or provide additional information

oai_token-act-pair · gpt-3.5-turbo

phrases addressing a specific audience or group of people

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

enegger

-0.74

forth

-0.67

ãģ®å

-0.67

Cho

-0.67

 Cheong

-0.65

Director

-0.65

orial

-0.64

 Discrimination

-0.64

Dialog

-0.63

maker

-0.61

POSITIVE LOGITS

 sake

1.21

 purposes

1.04

 wishing

0.86

 reasons

0.84

 curious

0.83

 redes

0.82

ummies

0.80

 unfamiliar

0.79

 indul

0.79

 interested

0.78

Activations Density 0.071%

phrases that introduce a new topic or provide additional information

phrases addressing a specific audience or group of people

No Comments

No Known Activations