INDEX

Explanations

.m/.hpp/.java/.cpp

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

uppercase acronyms and camelCase identifiers in programming code.

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

The neuron detects tokens consisting of multi‐letter uppercase sequences (acronyms or initialisms).

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

</h2>

-2.67

and

-2.31

 Instead

-1.82

</em>

-1.77

 which

-1.70

It

-1.64

Thus

-1.63

伯爵

-1.60

are

-1.59

so

-1.57

POSITIVE LOGITS

余談

1.98

 pracovní

1.77

あぁ

1.69

こんな感じで

1.68

áním

1.68

 povezave

1.68

因为他

1.63

 jejich

1.63

 různých

1.63

ごちそう

1.61

Activations Density 0.097%

.m/.hpp/.java/.cpp

uppercase acronyms and camelCase identifiers in programming code.

The neuron detects tokens consisting of multi‐letter uppercase sequences (acronyms or initialisms).

No Comments

No Known Activations

.m/.hpp/.java/.cpp

uppercase acronyms and camelCase identifiers in programming code.

The neuron detects tokens consisting of multi‐letter uppercase sequences (acronyms or initialisms).

No Comments

No Known Activations