INDEX

Explanations

mentions of things that are extremely high quality or advanced

oai_token-act-pair · gpt-3.5-turbo

the term "ultra" 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 @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

 Burnett

-0.73

Interstitial

-0.72

 Bleach

-0.71

 Laugh

-0.68

 Nights

-0.68

 Doct

-0.67

ë

-0.66

 Regener

-0.65

akespeare

-0.65

 Gott

-0.64

POSITIVE LOGITS

orthodox

0.97

 ultra

0.95

visor

0.95

charged

0.84

entimes

0.77

 ultr

0.75

avorite

0.74

iolet

0.73

ixtape

0.72

thodox

0.72

Activations Density 0.004%

mentions of things that are extremely high quality or advanced

the term "ultra" in various contexts

No Comments

No Known Activations