INDEX

Explanations

chair-related phrases and actions

oai_token-act-pair · gpt-3.5-turbo

references to various types of chairs

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

 Territories

-0.69

aries

-0.68

Adds

-0.65

actionDate

-0.64

arily

-0.64

 persever

-0.63

 Rounds

-0.63

ocious

-0.62

DN

-0.62

ï¸ı

-0.61

POSITIVE LOGITS

lift

1.37

holder

1.07

 chair

1.07

manship

1.04

 chairs

1.03

chair

1.01

tops

0.97

 cush

0.96

woman

0.94

chairs

0.94

Activations Density 0.025%

chair-related phrases and actions

references to various types of chairs

No Comments

No Known Activations