INDEX

Explanations

distinguishes, basic, thinly veiled

The neuron detects speculative or conditional phrasing—words and patterns used in “if… then” constructs, modals like “would” or “might,” and other cues signalling hypothesis or argument.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 délais

-1.43

 médec

-1.42

virons

-1.35

 broderie

-1.24

titudine

-1.23

 rayons

-1.22

人民币

-1.21

 réserver

-1.20

пкой

-1.20

orté

-1.16

POSITIVE LOGITS

 here

1.44

in

1.30

as

1.29

but

1.21

 into

1.18

 with

1.17

 различни

1.16

 responses

1.15

 might

1.13

 both

1.12

Activations Density 0.005%