INDEX

Explanations

words and phrases that indicate quotations or citations

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

numbers and punctuation followed by words

np_acts-logits-general · gemini-2.5-flash-lite

the neuron detects document-structure tokens and metadata markers (section/figure numbers, citation/heading tokens, and similar punctuation/symbol tokens).

oai_token-act-pair · gpt-5-mini Triggered by @vetterc0

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 9-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_9/width_131k/average_l0_121

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.9.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

<unused8>

-1.04

[@BOS@]

-1.04

<unused41>

-1.04

<unused68>

-1.04

<unused16>

-1.03

<unused14>

-1.03

<unused23>

-1.03

<unused28>

-1.03

<pad>

-1.03

<unused3>

-1.03

POSITIVE LOGITS

 elétrico

0.27

↵↵

0.26

 resultaat

0.25

 sekunder

0.24

 samym

0.24

 elástico

0.23

 actuales

0.23

 atuais

0.23

 resultaten

0.23

 działania

0.22

Activations Density 0.738%

words and phrases that indicate quotations or citations

numbers and punctuation followed by words

the neuron detects document-structure tokens and metadata markers (section/figure numbers, citation/heading tokens, and similar punctuation/symbol tokens).

No Comments

No Known Activations