INDEX

Explanations

REPL and command-line instructions

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

command-line interface prompts and output from programming tools like debuggers (gdb), interpreters (irb, Scala REPL), and consoles.

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

This neuron detects tokens characteristic of interactive REPL or debugger session prompts and metadata (e.g. “irb(main):…>”, “gdb command file”, etc.).

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

 córdoba

-1.42

 tinto

-1.38

 innan

-1.31

 furia

-1.28

 cristo

-1.27

 handlar

-1.27

 vinci

-1.27

 bomba

-1.25

 méxico

-1.24

 ching

-1.23

POSITIVE LOGITS

1.37

own

1.26

 obtiene

1.23

downcase

1.22

why

1.19

 verbessert

1.18

WHY

1.18

</h1>

1.16

("

1.15

Activations Density 0.060%

REPL and command-line instructions

command-line interface prompts and output from programming tools like debuggers (gdb), interpreters (irb, Scala REPL), and consoles.

This neuron detects tokens characteristic of interactive REPL or debugger session prompts and metadata (e.g. “irb(main):…>”, “gdb command file”, etc.).

No Comments

No Known Activations

REPL and command-line instructions

command-line interface prompts and output from programming tools like debuggers (gdb), interpreters (irb, Scala REPL), and consoles.

This neuron detects tokens characteristic of interactive REPL or debugger session prompts and metadata (e.g. “irb(main):…>”, “gdb command file”, etc.).

No Comments

No Known Activations