INDEX

Explanations

comment starter tokens

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

The neuron fires strongly on block‐comment markers and their contents (e.g. /**, /, and the leading “” lines), effectively identifying multi-line comment blocks (especially Javadoc-style comments).

oai_token-act-pair · o4-mini Triggered by @jyhe0408

New Auto-Interp

Configuration

google/gemma-scope-27b-pt-res/layer_34/width_131k

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

你来

-0.93

 Karam

-0.89

と思いました

-0.83

 Raceway

-0.81

 abſ

-0.81

jalá

-0.81

turns

-0.79

鰯

-0.79

 Thistle

-0.79

 gradi

-0.79

POSITIVE LOGITS

on

1.33

 before

1.20

 through

1.03

by

0.96

 time

0.94

 during

0.91

android

0.86

Rx

0.81

 will

0.81

 Deshalb

0.79

Activations Density 0.013%

comment starter tokens

The neuron fires strongly on block‐comment markers and their contents (e.g. /**, /, and the leading “” lines), effectively identifying multi-line comment blocks (especially Javadoc-style comments).

No Comments

No Known Activations

comment starter tokens

The neuron fires strongly on block‐comment markers and their contents (e.g. /**, /, and the leading “” lines), effectively identifying multi-line comment blocks (especially Javadoc-style comments).

No Comments

No Known Activations

comment starter tokens

The neuron fires strongly on block‐comment markers and their contents (e.g. /**, */, and the leading “*” lines), effectively identifying multi-line comment blocks (especially Javadoc-style comments).

No Comments

No Known Activations

comment starter tokens

The neuron fires strongly on block‐comment markers and their contents (e.g. /**, */, and the leading “*” lines), effectively identifying multi-line comment blocks (especially Javadoc-style comments).

No Comments

No Known Activations

The neuron fires strongly on block‐comment markers and their contents (e.g. /**, /, and the leading “” lines), effectively identifying multi-line comment blocks (especially Javadoc-style comments).

The neuron fires strongly on block‐comment markers and their contents (e.g. /**, /, and the leading “” lines), effectively identifying multi-line comment blocks (especially Javadoc-style comments).