INDEX

Explanations

interrogative phrases, particularly those beginning with "what," "how," and "why."

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

what ishow dowhat arecan we

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

paravant

-0.57

 quidem

-0.40

selves

-0.39

 túnel

-0.38

 paraître

-0.38

 lópez

-0.35

The

-0.35

 nämlich

-0.34

drawing

-0.34

 proprement

-0.33

POSITIVE LOGITS

 autorytatywna

0.68

 المعيارى

0.66

出版年

0.62

 onCancelled

0.60

 esternos

0.57

 فريبيس

0.57

хьтан

0.53



0.53

tvguidetime

0.53

 onBind

0.53

Activations Density 0.025%

interrogative phrases, particularly those beginning with "what," "how," and "why."

what ishow dowhat arecan we

No Comments

No Known Activations