INDEX

Explanations

improve health and function

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

verbs associated with improving, increasing, or reducing health outcomes in product claims or medical research.

oai_token-act-pair · claude-3-7-sonnet-20250219 Triggered by @neilrathi

The neuron fires on verbs that describe quantitative health or performance effects—i.e. words like “improve,” “increase,” “reduce,” “boost,” or “lower.”

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

to

-2.33

出了一

-1.53

-1.45

 mientras

-1.42

桫

-1.40

箔

-1.38

The

-1.37

oning

-1.36

gnore

-1.36

omsday

-1.34

POSITIVE LOGITS

Both

1.74

 both

1.67

 your

1.67

你

1.55

ないです

1.46

因為

1.45

ׇ

1.42

iedene

1.40

Without

1.37

要

1.36

Activations Density 0.099%

improve health and function

verbs associated with improving, increasing, or reducing health outcomes in product claims or medical research.

The neuron fires on verbs that describe quantitative health or performance effects—i.e. words like “improve,” “increase,” “reduce,” “boost,” or “lower.”

No Comments

No Known Activations

improve health and function

verbs associated with improving, increasing, or reducing health outcomes in product claims or medical research.

The neuron fires on verbs that describe quantitative health or performance effects—i.e. words like “improve,” “increase,” “reduce,” “boost,” or “lower.”

No Comments

No Known Activations