INDEX

Explanations

terms associated with programming or machine learning concepts

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

pleas and legal proceedings

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

Descripción

-0.36

 igång

-0.36

건

-0.35

 ervan

-0.34

ombic

-0.34

 répartition

-0.34

 shots

-0.33

eseorang

-0.33

 rund

-0.32

sch

-0.32

POSITIVE LOGITS

 незавершена

0.61

jmniej

0.52

而是

0.52

EDEFAULT

0.49

rungsseite

0.49

 AspNetCore

0.49

TagMode

0.49

AndEndTag

0.47

 بلکه

0.47

 חיצוניים

0.47

Activations Density 0.102%

terms associated with programming or machine learning concepts

pleas and legal proceedings

No Comments

No Known Activations