INDEX

Explanations

phrases related to news or journalism

oai_token-act-pair · gpt-3.5-turbo

titles or phrases that suggest important or notable topics

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ittal

-0.73

ioned

-0.67

oubted

-0.64

lie

-0.60

ritten

-0.60

ibly

-0.59

>[

-0.59

munition

-0.58

 pocket

-0.57

 flushed

-0.56

POSITIVE LOGITS

Ltd

0.82

 Excellence

0.78

 Seasons

0.78

 Legends

0.78

 Enhancement

0.77

 Humanity

0.76

 Society

0.74

 Programming

0.73

 Madness

0.72

 Lights

0.72

Activations Density 1.801%

phrases related to news or journalism

titles or phrases that suggest important or notable topics

No Comments

No Known Activations