INDEX

Explanations

terms related to utilities or services

oai_token-act-pair · gpt-3.5-turbo

references to utility in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Spielberg

-0.73

 rehears

-0.69

aughs

-0.68

 Kinn

-0.68

ceans

-0.68

 Starr

-0.68

sth

-0.67

raised

-0.65

jing

-0.65

heard

-0.65

POSITIVE LOGITS

 utility

4.00

 Utility

2.91

 utilities

2.74

 Utilities

2.03

Ut

1.43

 usefulness

1.36

 utilitarian

1.30

 convenience

1.13

 appliance

1.11

 applic

1.08

Activations Density 0.017%

terms related to utilities or services

references to utility in various contexts

No Comments

No Known Activations