INDEX

Explanations

terms related to potential identification or classification

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

refer to definitions

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

 manifestly

-0.36

 consistently

-0.35

 koment

-0.34

empêcher

-0.33

 continuously

-0.33

const

-0.32

 quase

-0.32

 Regel

-0.32

 okam

-0.32

 instructed

-0.32

POSITIVE LOGITS

ValueStyle

0.87

 незавершена

0.71

хьтан

0.71

 doInBackground

0.70

oa̍t

0.70

 مرئيه

0.69

 vPvB

0.68

 ujednoznacz

0.68

 հղումներ

0.64

 يتيمه

0.64

Activations Density 0.000%

terms related to potential identification or classification

refer to definitions

No Comments

No Known Activations