INDEX

Explanations

references to large and massive physical objects

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

giant size

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

 chrétien

-0.36

 mukana

-0.36

före

-0.35

 monasterio

-0.34

 foncé

-0.34

 aislados

-0.34

 competencias

-0.33

 Tiefen

-0.32

 conseguenza

-0.32

 passou

-0.32

POSITIVE LOGITS

 gigantic

0.79

 giant

0.74

XXL

0.68

Giant

0.66

lossal

0.66

 Giant

0.66

giant

0.64

 colossal

0.63

verwijspagina

0.61

 gigantes

0.59

Activations Density 0.019%

references to large and massive physical objects

giant size

No Comments

No Known Activations