INDEX

Explanations

geometric shapes, particularly ovals and ellipses

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

 fédé

-0.48

 démocr

-0.46

NavController

-0.45

Schedulers

-0.44

🥞

-0.43

dur

-0.42

ORI

-0.40

 język

-0.40

Dur

-0.40

 Thermodynamics

-0.40

POSITIVE LOGITS

 oval

1.02

 Oval

1.00

Oval

0.96

ellipse

0.85

oval

0.82

 ovale

0.81

Ellipse

0.79

 ellipse

0.78

 Ellipse

0.75

RectF

0.67

Activations Density 0.011%

geometric shapes, particularly ovals and ellipses

No Comments

No Known Activations