INDEX

Explanations

references to television in various contexts

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

television broadcasting

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

 Aussie

-0.50

quote

-0.47

Mighty

-0.45

Buat

-0.44

boost

-0.43

arg

-0.43

 Combo

-0.43

split

-0.43

ugget

-0.43

booking

-0.42

POSITIVE LOGITS

 Television

1.37

 television

1.24

Television

1.23

television

1.20

 televisions

1.00

 télévision

0.98

 televisão

0.91

 televis

0.90

tv

0.87

 televisión

0.87

Activations Density 0.005%

references to television in various contexts

television broadcasting

No Comments

No Known Activations