INDEX

Explanations

reporting speech

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

The neuron activates on speaker-attribution phrases, i.e. proper names or pronouns coupled with reporting verbs like “said,” “explained,” or “added.”

oai_token-act-pair · o4-mini Triggered by @jyhe0408

journalistic attribution clauses indicating who is speaking, typically a pronoun or name followed by a reporting verb.

oai_token-act-pair · gpt-5 Triggered by @jyhe0408

quotations or reported speech, particularly the word "said" when attributing statements to a speaker.

oai_token-act-pair · claude-4-5-sonnet 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

he

-1.58

-1.30

manchester

-1.29

``

-1.27

]);

-1.26

â

-1.26

icksburg

-1.23

 Američ

-1.22

谨慎

-1.20

ródła

-1.20

POSITIVE LOGITS

of

1.35

话音

1.33

for

1.23

 then

1.19

 WHEREAS

1.14

 soulign

1.13

뺀

1.13

ção

1.13

Activations Density 0.012%

reporting speech

The neuron activates on speaker-attribution phrases, i.e. proper names or pronouns coupled with reporting verbs like “said,” “explained,” or “added.”

journalistic attribution clauses indicating who is speaking, typically a pronoun or name followed by a reporting verb.

quotations or reported speech, particularly the word "said" when attributing statements to a speaker.

No Comments

No Known Activations

reporting speech

The neuron activates on speaker-attribution phrases, i.e. proper names or pronouns coupled with reporting verbs like “said,” “explained,” or “added.”

journalistic attribution clauses indicating who is speaking, typically a pronoun or name followed by a reporting verb.

quotations or reported speech, particularly the word "said" when attributing statements to a speaker.

No Comments

No Known Activations