INDEX

Explanations

programming constructs and code-related elements

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

Code fragments setting box width values

np_acts-logits-general · gemini-2.0-flash

newMax<end_of_turn>

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-16k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_16k/average_l0_91

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

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

multer

-0.44

AndEndTag

-0.43

 potencial

-0.43

 Ground

-0.41

 business

-0.41

 consum

-0.40

supset

-0.40

business

-0.39

 Plut

-0.39

 Hypothe

-0.39

POSITIVE LOGITS

 ulang

0.48

 anew

0.46

ElementException

0.45



0.44

 betale

0.43

 igjen

0.42

onResume

0.40

 animación

0.40

 nových

0.39

 arvo

0.39

Activations Density 0.083%

programming constructs and code-related elements

Code fragments setting box width values

newMax<end_of_turn>

No Comments

No Known Activations