INDEX

Explanations

terms related to mathematical coefficients and relationships in calibration and modeling contexts

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

five justices

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

ukone

-0.54

 Overnight

-0.42

GraphicsUnit

-0.42

 overnight

-0.41

 FontWeight

-0.40

 nakalista

-0.39

༘

-0.39

Overnight

-0.39

!*\

-0.39

 ModelRenderer

-0.39

POSITIVE LOGITS

meetings

0.42

 meetings

0.42

 насељу

0.42

 فريبيس

0.38

archiviato

0.37

Meetings

0.37

meeting

0.36

 reunión

0.36

 meeting

0.36

EnableWeb

0.36

Activations Density 0.122%

terms related to mathematical coefficients and relationships in calibration and modeling contexts

five justices

No Comments

No Known Activations