INDEX

Explanations

unordered steps or instructions

oai_token-act-pair · gpt-3.5-turbo

procedural steps or instructions

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

 flourishing

-0.79

ciating

-0.78

 hearing

-0.70

 velvet

-0.68

lit

-0.68

 booming

-0.67

 balloon

-0.67

 balloons

-0.66

vag

-0.65

 smoke

-0.64

POSITIVE LOGITS

Step

1.41

hens

1.05

 Steps

1.05

step

0.96

STEP

0.94

 Step

0.87

 STEP

0.85

DragonMagazine

0.85

Mania

0.84

hent

0.84

Activations Density 0.007%

unordered steps or instructions

procedural steps or instructions

No Comments

No Known Activations