INDEX

Explanations

terms related to software APIs and frameworks

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

programming namespaces and classes

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

 olvidar

-0.50

 comunicarse

-0.49

 aveug

-0.47

 olvid

-0.45

 kemajuan

-0.44

 cuantos

-0.44

 gostar

-0.43

 capaces

-0.43

 cuánto

-0.43

 respetar

-0.43

POSITIVE LOGITS

 jadx

0.61

 betweenstory

0.60

المناصب

0.55

Попис

0.53

 appa

0.52

 bassist

0.52

 lira

0.51

 Chai

0.50

Ayr

0.49

ARP

0.49

Activations Density 0.054%

terms related to software APIs and frameworks

programming namespaces and classes

No Comments

No Known Activations