INDEX

Explanations

terms related to a wide range of a particular characteristic or concept, possibly representing a spectrum of that characteristic

oai_token-act-pair · gpt-3.5-turbo

references to various types of spectra

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

 Boss

-0.76

ador

-0.70

 Roose

-0.68

tto

-0.67

essors

-0.66

abo

-0.66

Bai

-0.64

Jo

-0.63

keep

-0.63

jo

-0.62

POSITIVE LOGITS

 spectrum

4.18

 Spectrum

2.53

 continuum

2.17

 wavelength

1.34

Spect

1.29

 rainbow

1.27

 range

1.21

 wavelengths

1.17

spect

1.17

 spect

1.17

Activations Density 0.021%

terms related to a wide range of a particular characteristic or concept, possibly representing a spectrum of that characteristic

references to various types of spectra

No Comments

No Known Activations