INDEX

Explanations

negative numeric values

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

 pleaſure

-0.55

 Chwiliwch

-0.50

 houſe

-0.47

 disfraces

-0.46

 nahilalakip

-0.46

 romántico

-0.45

出版年

-0.45

Gön

-0.45

verwijspagina

-0.44

 Autorisations

-0.43

POSITIVE LOGITS

encodeWith

0.60

mtrl

0.49

:✨

0.46

waitKey

0.42

webpack

0.42

unjungi

0.41

Vidite

0.39

EDEFAULT

0.39

Abitanti

0.39

Попис

0.39

Activations Density 0.007%

negative numeric values

No Comments

No Known Activations