INDEX

Explanations

phrases indicating uncertainty or questions related to validation and compliance

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

foreign or negation

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

ChildScrollView

-0.54

 CanadaChoose

-0.48

賀状

-0.48

ThroughAttribute

-0.45

igte

-0.44

참고

-0.44

kegaard

-0.43

 Numerade

-0.43

出版年

-0.42

âce

-0.42

POSITIVE LOGITS

 altrimenti

0.38

flashdata

0.36

 otherwise

0.36

 sinon

0.33

cant

0.33

zil

0.33

foreign

0.33

 NUKAT

0.33

 cant

0.32

Cant

0.32

Activations Density 0.108%

phrases indicating uncertainty or questions related to validation and compliance

foreign or negation

No Comments

No Known Activations