INDEX

Explanations

a focus on significant numerical values or high activation counts in various contexts

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

years or dates

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

Autoritní

-0.60

 zoude

-0.56

 мәкал

-0.54

 Chwiliwch

-0.54

 dezelve

-0.52

SourceChecksum

-0.52

tagHelperRunner

-0.51

يكب

-0.49

 leão

-0.49

 zelve

-0.49

POSITIVE LOGITS

getMock

0.47

↵↵

0.44

Clik

0.44

strijden

0.42

↵

0.40

datastore

0.39

spre

0.38

vor

0.37

 manner

0.37

mtrl

0.37

Activations Density 0.017%

a focus on significant numerical values or high activation counts in various contexts

years or dates

No Comments

No Known Activations