INDEX

Explanations

parentheses and brackets in code-related contexts

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

( followed by specific characters

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_24/width_131k/average_l0_97

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.24.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

 ſta

-0.46

 paſſ

-0.46

 ſtre

-0.45

 juſ

-0.45

 tranſ

-0.45

 preſent

-0.44

MigrationBuilder

-0.43

 viſ

-0.40

 Inſ

-0.40

 ſol

-0.39

POSITIVE LOGITS

+#+#

0.57

 jScrollPane

0.56

sizeCache

0.56

awtextra

0.54

 initComponents

0.53

#(

0.52

三四

0.52

----</

0.52

 CanadaChoose

0.51

 }^{(

0.49

Activations Density 0.031%

parentheses and brackets in code-related contexts

( followed by specific characters

No Comments

No Known Activations