INDEX

Explanations

terms related to visibility and status indicators

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

absence or negation

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

 يتيمه

-0.54

 propOrder

-0.49

hyrchwyd

-0.49

+#+#

-0.48

ArrowToggle

-0.48

fortawesome

-0.46

:+:

-0.45

StoryboardSegue

-0.45

☸

-0.45

InitVars

-0.45

POSITIVE LOGITS

 hingegen

0.43

 wikipagina

0.40

fjspx

0.40

 verschillen

0.40

GOTREF

0.39

Externé

0.35

 gedrag

0.35

xffffffff

0.34

Israël

0.33

 seduta

0.32

Activations Density 0.109%

terms related to visibility and status indicators

absence or negation

No Comments

No Known Activations