INDEX

Explanations

comparative phrases and words indicating quantity or measurement

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

quantities and multipliers

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 maupun

-0.43

 postup

-0.37

 arrangement

-0.32

 lecz

-0.32

 Anordnung

-0.31

ist

-0.30

 arranged

-0.30

하며

-0.30

 Botschaft

-0.30

 Versammlung

-0.30

POSITIVE LOGITS

 мәкал

0.73

GEBURTSDATUM

0.72

 Wikimedijinoj

0.71

 invokingState

0.69

<unused14>

0.69

<unused52>

0.69

$_(

0.69

<unused28>

0.69

<unused8>

0.69

<unused16>

0.69

Activations Density 0.030%

comparative phrases and words indicating quantity or measurement

quantities and multipliers

No Comments

No Known Activations