INDEX

Explanations

phrases related to structural components and layered configurations in scientific contexts

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

spatial or positional references

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

DeleteBehavior

-0.56

 فريبيس

-0.49

HomeAsUpEnabled

-0.45

össe

-0.44

 Réalisation

-0.43

 exercice

-0.42

 addSubview

-0.41

endcsname

-0.41

classifiers

-0.40

 Référence

-0.40

POSITIVE LOGITS

 adjacent

0.47

adjacent

0.45

፩

0.41

 surrounding

0.41

 dietro

0.41

 behind

0.40

 الاطلاع

0.40

 bên

0.40

 voorbij

0.40

Hentet

0.40

Activations Density 0.149%

phrases related to structural components and layered configurations in scientific contexts

spatial or positional references

No Comments

No Known Activations