INDEX

Explanations

lists of items, such as countries, representatives, or movies

oai_token-act-pair · gpt-3.5-turbo

phrases indicating lists

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

ester

-0.72

entimes

-0.72

imet

-0.71

gypt

-0.69

iva

-0.67

imeter

-0.65

 breeze

-0.64

athy

-0.64

lycer

-0.63

ashtra

-0.62

POSITIVE LOGITS

 sorts

0.89

 entries

0.85

 accomplishments

0.79

ãĥ¼ãĥ³

0.76

 items

0.75

Topics

0.74

 ingredients

0.72

 celebrities

0.71

 keywords

0.71

 grievances

0.70

Activations Density 0.107%

lists of items, such as countries, representatives, or movies

phrases indicating lists

No Comments

No Known Activations