INDEX

Explanations

code-related constructs and logical conditions in programming

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

sports abbreviations

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

舺

-0.54

 Exacts

-0.52

 gyhoeddwyd

-0.50

 مرئيه

-0.50

 Италијани

-0.49

 AssemblyCulture

-0.48

XmlAccessType

-0.47

олові

-0.47

تقاوى

-0.47

ViewFeatures

-0.46

POSITIVE LOGITS

 middle

0.36

Kob

0.35

 intermediate

0.34

 intervening

0.34

 beteiligt

0.33

Rog

0.31

 farinha

0.31

Pet

0.31

op

0.31

ngOnInit

0.30

Activations Density 0.557%

code-related constructs and logical conditions in programming

sports abbreviations

No Comments

No Known Activations