INDEX

Explanations

specifically labeled sections within a document, most likely related to full information or content

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Full" indicating complete transcripts or articles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Downloadha

-0.78

ovan

-0.78

åĤ

-0.75

*/(

-0.74

estern

-0.72

 Michaels

-0.70

pton

-0.69

oice

-0.69

Wars

-0.67

senal

-0.66

POSITIVE LOGITS

erton

1.04

screen

0.87

bright

0.81

 fled

0.76

ening

0.75

agar

0.74

ness

0.74

ature

0.72

 frontal

0.72

desktop

0.72

Activations Density 0.021%

specifically labeled sections within a document, most likely related to full information or content

instances of the word "Full" indicating complete transcripts or articles

No Comments

No Known Activations