INDEX

Explanations

version numbers

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

version numbers in semantic versioning format (e.g., v1.0.0, v0.8.1).

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

The neuron selectively fires on numeric tokens—especially multi‐digit numbers and decimal points—i.e. version numbers or other floating‐point values.

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

犼

-1.80

 пришла

-1.66

橚

-1.64

gunas

-1.63

⧼

-1.63

黻

-1.57

熠

-1.55

 begleitet

-1.54

啬

-1.52

-1.49

POSITIVE LOGITS

;"

1.55

BIÉN

1.52

眼光

1.52

1.48

 как

1.41

?"

1.41

);

1.41

ânsito

1.41

殃

1.40

"'

1.38

Activations Density 0.060%

version numbers

version numbers in semantic versioning format (e.g., v1.0.0, v0.8.1).

The neuron selectively fires on numeric tokens—especially multi‐digit numbers and decimal points—i.e. version numbers or other floating‐point values.

No Comments

No Known Activations

version numbers

version numbers in semantic versioning format (e.g., v1.0.0, v0.8.1).

The neuron selectively fires on numeric tokens—especially multi‐digit numbers and decimal points—i.e. version numbers or other floating‐point values.

No Comments

No Known Activations