INDEX

Explanations

specific words and structures related to processes and methods in technical or scientific contexts

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

method applied to materials

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

wortet

-0.36

why

-0.33

 otros

-0.33

How

-0.32

#+#

-0.32

-0.31

Why

-0.31

How

-0.30

Mar

-0.30

?.

-0.30

POSITIVE LOGITS

ValueStyle

0.67

ⓧ

0.64

ImageContext

0.61

 Jefus

0.59

 houſe

0.59

 ddelwed

0.58

 cherchés

0.57

yarnpkg

0.56

 perſon

0.56

SequentialGroup

0.55

Activations Density 0.139%

specific words and structures related to processes and methods in technical or scientific contexts

method applied to materials

No Comments

No Known Activations