INDEX

Explanations

variable assignment

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

single-letter variable names in code.

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

The neuron strongly activates on numeric literals (and variable tokens whose activation scales with the numeric value they’re assigned), i.e. it detects numbers in code.

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

 jesteśmy

-1.52

 gewähr

-1.48

 oglą

-1.38

non

-1.38

 almak

-1.37

 lequel

-1.34

號

-1.33

import

-1.30

nen

-1.30

About

-1.28

POSITIVE LOGITS

 with

1.79

 other

1.61

for

1.60

 because

1.45

 that

1.41

Ţ

1.40

 andere

1.34

 anderen

1.30

всем

1.29

 diğer

1.28

Activations Density 0.093%

variable assignment

single-letter variable names in code.

The neuron strongly activates on numeric literals (and variable tokens whose activation scales with the numeric value they’re assigned), i.e. it detects numbers in code.

No Comments

No Known Activations

variable assignment

single-letter variable names in code.

The neuron strongly activates on numeric literals (and variable tokens whose activation scales with the numeric value they’re assigned), i.e. it detects numbers in code.

No Comments

No Known Activations