INDEX

Explanations

phrases related to criminal allegations and misconduct

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

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 <<<<<<<<<<<<<<

-0.57

Autoritní

-0.56

 autorytatywna

-0.55

енча

-0.47

usermodel

-0.46

الإنجليزية

-0.44

 efficaces

-0.41

 šal

-0.41

 palk

-0.40

 DEPOSITORY

-0.40

POSITIVE LOGITS

 scandals

0.65

 allegations

0.59

 scandal

0.52

 accusations

0.47

 concerns

0.47

 crimes

0.45

VersionUID

0.44

WritableDatabase

0.43

 հղումներ

0.43

 taint

0.43

Activations Density 0.061%

phrases related to criminal allegations and misconduct

No Comments

No Known Activations