INDEX

Explanations

phrases indicating proximity or adjacency

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

beside

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

 Cár

-0.47

 Agung

-0.44

 getDate

-0.42

 Hukum

-0.42

ungsbedingungen

-0.41

Eksteraj

-0.40

getFullYear

-0.40

 Warna

-0.40

setContentType

-0.39

prüche

-0.39

POSITIVE LOGITS

 beside

0.85

 Beside

0.79

 accanto

0.74

 obok

0.72

 neben

0.69

 рядом

0.66

 fianco

0.64

 neighbour

0.63

 naast

0.63

 возле

0.60

Activations Density 0.009%

phrases indicating proximity or adjacency

beside

No Comments

No Known Activations