INDEX

Explanations

numerical values and their associated units or identifiers

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

dimensions and resolutions

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

 ancaman

-0.49

 gekomen

-0.46

 gevolg

-0.45

 liitty

-0.44

 advertencia

-0.44

vů

-0.43

esistenza

-0.43

 uporab

-0.43

 insuffisamment

-0.42

 temprana

-0.42

POSITIVE LOGITS

BagLayout

0.55

XF

0.52

 getX

0.51

KX

0.49

 horizontal

0.49

 rectangle

0.48

AXB

0.48

XE

0.48

Ã

0.48

KX

0.48

Activations Density 0.035%

numerical values and their associated units or identifiers

dimensions and resolutions

No Comments

No Known Activations