INDEX

Explanations

specific characters or symbols from diverse languages and contexts

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

Japanese names and compound words

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

featureID

-0.66

 disambiguazione

-0.61

 ujednoznacz

-0.58

queryInterface

-0.57

EDEFAULT

-0.55

 propOrder

-0.52

verifyException

-0.51

StoryboardSegue

-0.50

@"/

-0.49

 mergeFrom

-0.48

POSITIVE LOGITS

prüche

0.45

 German

0.38

 Cyril

0.38

 pathlib

0.37

isinstance

0.37

Polish

0.36

andoah

0.36

German

0.36

 Polish

0.36

orological

0.36

Activations Density 0.015%

specific characters or symbols from diverse languages and contexts

Japanese names and compound words

No Comments

No Known Activations