INDEX

Explanations

elements related to specific names, terms, or categories in various contexts

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

Fuller, Jack, Etaoin shrdlu

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

 chrétiens

-0.44

 Gestión

-0.43

 Kriege

-0.43

 Besar

-0.41

 competitivo

-0.41

 récentes

-0.40

 først

-0.40

 publiques

-0.40

 entière

-0.39

 impuestos

-0.39

POSITIVE LOGITS

interopRequire

0.70

]='\

0.69

 AssemblyCulture

0.67

uxxxx

0.66

脚注の使い方

0.66

TacToe

0.63

NameInMap

0.60

copg

0.60

iſter

0.59

-------

0.59

Activations Density 0.184%

elements related to specific names, terms, or categories in various contexts

Fuller, Jack, Etaoin shrdlu

No Comments

No Known Activations