INDEX

Explanations

the word "pret" or variations of it, like "pretend" or "pretense"

oai_token-act-pair · gpt-3.5-turbo

mentions of pretzels and related word forms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

-0.75

SHIP

-0.65

 Leopard

-0.65

 Sierra

-0.64

 Blessed

-0.63

 Feed

-0.63

uador

-0.60

 Bullet

-0.60

 Moving

-0.60

 Patriarch

-0.60

POSITIVE LOGITS

 pret

1.27

zel

0.98

ensions

0.96

entious

0.94

anooga

0.87

ension

0.87

reatment

0.86

inen

0.86

pret

0.81

ional

0.80

Activations Density 0.005%

the word "pret" or variations of it, like "pretend" or "pretense"

mentions of pretzels and related word forms

No Comments

No Known Activations