INDEX

Explanations

terms related to the process of refining or improvement

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

making things better

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

 trá

-0.46

writeValue

-0.46

 Joey

-0.43

BrowserModule

-0.43

 <>",

-0.43

 raven

-0.42

vertx

-0.42

velocityX

-0.42

adaptiveStyles

-0.42

Joey

-0.41

POSITIVE LOGITS

 refine

0.81

 honed

0.79

 refined

0.79

refined

0.78

 refinement

0.75

 refining

0.73

refine

0.72

 perfected

0.71

 refinements

0.71

Perfection

0.68

Activations Density 0.022%

terms related to the process of refining or improvement

making things better

No Comments

No Known Activations