INDEX

Explanations

phrases that indicate restrictions or limitations in legal contexts

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

changing or overriding decisions

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

 CanadaChoose

-0.65

labus

-0.46

 Result

-0.46

.*")]

-0.44

 OMITBAD

-0.41

MessageTagHelper

-0.41

Aiheesta

-0.41

 AssemblyVersion

-0.41

 timed

-0.41

 Privacidade

-0.41

POSITIVE LOGITS

 kuiten

0.45

 برانيه

0.42

 nicio

0.42

 bolsillo

0.40

oules

0.40

SharedDtor

0.40

 jemals

0.39

 kaupung

0.38

ynarod

0.37

berdayakan

0.37

Activations Density 0.107%

phrases that indicate restrictions or limitations in legal contexts

changing or overriding decisions

No Comments

No Known Activations