INDEX

Explanations

programming input/output

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

code reading user input from standard input or console.

oai_token-act-pair · claude-4-5-sonnet Triggered by @bfgcn7frf5

The neuron fires on occurrences of Java’s Scanner-based input‐reading API (e.g. “Scanner,” “nextInt,” “readLine,” and “in”).

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

New Auto-Interp

Configuration

google/gemma-scope-27b-pt-res/layer_10/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

Alternatively

-1.18

مانية

-1.17

щего

-1.16

 operações

-1.14

ógico

-1.13

hilterra

-1.12

IMHO

-1.12

͡

-1.12

 müssen

-1.11

さの

-1.11

POSITIVE LOGITS

 help

1.40

<td>

1.29

匙

1.28

淠

1.27

 wcześniej

1.25

={

1.25

 amounts

1.22

'].

1.20

 betrifft

1.20

 besonderes

1.19

Activations Density 0.018%

programming input/output

code reading user input from standard input or console.

The neuron fires on occurrences of Java’s Scanner-based input‐reading API (e.g. “Scanner,” “nextInt,” “readLine,” and “in”).

No Comments

No Known Activations

programming input/output

code reading user input from standard input or console.

The neuron fires on occurrences of Java’s Scanner-based input‐reading API (e.g. “Scanner,” “nextInt,” “readLine,” and “in”).

No Comments

No Known Activations