INDEX

Explanations

technical terms related to form and design

oai_token-act-pair · gpt-3.5-turbo

references to different types of forms or formats

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 Hots

-0.64

 nephew

-0.62

 spree

-0.61

 unsus

-0.60

 Silence

-0.58

 bang

-0.57

dry

-0.57

âĶĢâĶĢ

-0.56

 railing

-0.55

 Throne

-0.55

POSITIVE LOGITS

aldehyde

1.64

idable

1.47

ulating

1.39

ative

1.39

atted

1.34

ulators

1.33

atter

1.31

ulator

1.30

ality

1.23

ulas

1.23

Activations Density 0.048%

technical terms related to form and design

references to different types of forms or formats

No Comments

No Known Activations