INDEX

Explanations

structural elements in code snippets

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

conditional code

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

الدراسه

-0.91

ſelves

-0.72

 للمعارف

-0.70

ſelf

-0.69

leſs

-0.68

 ſch

-0.68

 Numerade

-0.67

 pleaſure

-0.65

 الرياضيه

-0.65

 huawei

-0.65

POSITIVE LOGITS

The

0.59

the

0.54

The

0.47

 gelangen

0.45

 Osborn

0.43

On

0.42

 Antrag

0.42

if

0.41

0.40

 Folge

0.39

Activations Density 0.011%

structural elements in code snippets

conditional code

No Comments

No Known Activations