INDEX

Explanations

words related to heights or levels of something

oai_token-act-pair · gpt-3.5-turbo

references to highness or elevation

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ISO

-0.77

 Machina

-0.73

 Bubble

-0.73

ahime

-0.70

 champagne

-0.70

Osc

-0.65

nesia

-0.64

 Gutenberg

-0.62

é¾įå

-0.62

 Hades

-0.61

POSITIVE LOGITS

ighed

0.83

bors

0.81

atures

0.79

entimes

0.78

igh

0.77

teen

0.77

lean

0.76

ress

0.75

bor

0.75

«

0.74

Activations Density 0.007%

words related to heights or levels of something

references to highness or elevation

No Comments

No Known Activations