INDEX

Explanations

pronouns and references to individuals in various contexts

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

action or description following a word

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

ValueStyle

-0.56

guiente

-0.35

 Kochen

-0.35

 fometimes

-0.34

penuhi

-0.33

aurait

-0.32

 grecque

-0.32

億円

-0.32

 fallait

-0.32

 raiſ

-0.32

POSITIVE LOGITS

 ब्रेकडाउन

0.75

 ProtoMessage

0.60

 CanadaChoose

0.54

Personendaten

0.52

scribers

0.52

μβρίου

0.51

EXTERN

0.50

msgSender

0.50

Népesség

0.49

transQ

0.48

Activations Density 0.027%

pronouns and references to individuals in various contexts

action or description following a word

No Comments

No Known Activations