INDEX

Explanations

terms related to rankings or positions in lists

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

comparisons and exceptions

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

 puesta

-0.42

 jurídico

-0.38

 fría

-0.37

 problemática

-0.36

 entrado

-0.35

 templado

-0.35

 productivo

-0.35

 couvre

-0.35

 manifestación

-0.35

 empate

-0.35

POSITIVE LOGITS

IndentedString

0.84

脚注の使い方

0.67

 виправивши

0.66

 Италијани

0.65

ValueStyle

0.64

 فريبيس

0.64

الإنجليزية

0.64

OGND

0.61

bkz

0.60

featureID

0.60

Activations Density 0.011%

terms related to rankings or positions in lists

comparisons and exceptions

No Comments

No Known Activations