INDEX

Explanations

instances of definite articles and specific phrases that indicate a focus on common subjects or themes in context

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

the place or event names

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

 chi̍t

-0.39

 zimní

-0.32

 gemeenten

-0.31

principalColumn

-0.30

 gezien

-0.29

 jezelf

-0.29

 retos

-0.28

 tornillos

-0.27

 Zukunft

-0.27

 trouw

-0.26

POSITIVE LOGITS

はじめに

0.81

 Autorizaciones

0.68

Jeografia

0.66

 Signalez

0.64

complexContent

0.62

يكب

0.61

ineſs

0.60

 noDo

0.60

MigrationBuilder

0.59

IUrlHelper

0.59

Activations Density 0.007%

instances of definite articles and specific phrases that indicate a focus on common subjects or themes in context

the place or event names

No Comments

No Known Activations