INDEX

Explanations

the start of a sentence or question, particularly focusing on words like "Which", "How", "Question", or other sentence-initial words.

oai_token-act-pair · claude-3-5-sonnet-20240620 Triggered by @myinzi123

keywords and phrases indicative of technical or formal content, particularly in programming or legal contexts

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

the first word or words at the beginning of a new document or paragraph, often names, greetings, or introductory phrases.

oai_token-act-pair · gpt-4.1-2025-04-14 Triggered by @ywhp2c8tjm

the beginning of a text.

oai_token-act-pair · gemini-2.0-flash Triggered by @ywhp2c8tjm

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-16k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_16k/average_l0_91

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 divino

-0.25

 kirke

-0.24

 […]

-0.24

 bougies

-0.23

 chien

-0.23

 trous

-0.23

 rire

-0.22

 peuple

-0.22

-0.21

 trés

-0.21

POSITIVE LOGITS

 Савезне

1.68

SourceChecksum

1.16

Personendaten

1.14

:✨

1.10

webElementXpaths

1.07

 Italijanski

1.05

 мәкал

1.04

 Biôgrafia

0.99

 Италијани

0.99

 Мексичка

0.98

Activations Density 5.209%

the start of a sentence or question, particularly focusing on words like "Which", "How", "Question", or other sentence-initial words.

keywords and phrases indicative of technical or formal content, particularly in programming or legal contexts

the first word or words at the beginning of a new document or paragraph, often names, greetings, or introductory phrases.

the beginning of a text.

No Comments

No Known Activations