INDEX

Explanations

specific articles and prepositions that indicate connection and relationship in discourse

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

more, no, can, wants

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

Personendaten

-0.46

новништво

-0.46

Autoritní

-0.44

 EconPapers

-0.40

 médec

-0.40

UnusedPrivate

-0.39

hoeddwyd

-0.38

 Bourgoin

-0.38

Referencer

-0.37

얻

-0.37

POSITIVE LOGITS

NameInMap

0.49

etc

0.49

etc

0.49

ついでに

0.47

先に

0.47

obenz

0.46

それでも

0.46

 even

0.46

IntoConstraints

0.46

 Crespo

0.45

Activations Density 0.050%

specific articles and prepositions that indicate connection and relationship in discourse

more, no, can, wants

No Comments

No Known Activations