INDEX

Explanations

instances of indexing and mathematical operations in code-like structures

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

j followed by math or code index

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

abstractmethod

-0.57

*}

-0.55

 itſelf

-0.54

__":

-0.54

بوابة

-0.54

【

-0.53

 BrowserModule

-0.53

 ..."

-0.52

 "*",

-0.51

="">

-0.50

POSITIVE LOGITS

0.84

0.79

0.76

0.65

JKLM

0.60

 jTextField

0.56

Jio

0.53

ji

0.53

 Jupyter

0.52

JI

0.52

Activations Density 0.108%

instances of indexing and mathematical operations in code-like structures

j followed by math or code index

No Comments

No Known Activations