INDEX

Explanations

references to replacement or new items

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

new replacement engines

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

wußt

-0.38

ElementException

-0.36

 Foot

-0.35

中海

-0.35

rumah

-0.34

ANTES

-0.34

ฤษ

-0.34

iterranean

-0.34

 BoxFit

-0.34

stylers

-0.34

POSITIVE LOGITS

replacement

0.66

 replacement

0.61

 replacements

0.55

 replacing

0.53

Replacement

0.52

 repuesto

0.52

replacing

0.52

 reemplazo

0.52

 remplacement

0.50

 Replacement

0.47

Activations Density 0.029%

references to replacement or new items

new replacement engines

No Comments

No Known Activations