INDEX

Explanations

terms related to architectural features and development potential

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

 compét

-0.35

 umane

-0.34

 suele

-0.31

 noël

-0.31

 Diretto

-0.31

 Głów

-0.31

 concernés

-0.30

 dictadura

-0.30

 computadoras

-0.30

 ministros

-0.30

POSITIVE LOGITS

IUrlHelper

0.76

 podium

0.65

DockStyle

0.57

 CommonModule

0.57

 tower

0.55

ExtendWith

0.54

 Urbana

0.53

󠁢

0.53

 absten

0.53

SequentialGroup

0.53

Activations Density 0.014%

terms related to architectural features and development potential

No Comments

No Known Activations