INDEX

Explanations

names ending in "eto"

oai_token-act-pair · gpt-3.5-turbo

names of specific diets or dietary plans

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

Ort

-0.71

 faults

-0.71

 Lazarus

-0.69

Pax

-0.67

nings

-0.67

Topics

-0.66

teenth

-0.66

 perceptions

-0.66

lies

-0.66

dn

-0.66

POSITIVE LOGITS

chnology

0.91

eto

0.90

veyard

0.87

Redditor

0.86

uesday

0.84

zzi

0.84

avascript

0.83

ourt

0.83

utonium

0.80

hedral

0.79

Activations Density 0.009%

names ending in "eto"

names of specific diets or dietary plans

No Comments

No Known Activations