INDEX

Explanations

yet... yet.

The neuron detects contrastive or concessive discourse cues that signal a shift or counterpoint in the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

但是

-1.27

 also

-1.27

but

-1.15

 però

-1.09

 certainly

-1.08

 imidlertid

-1.05

 especially

-1.05

 nhưng

-1.02

 definitely

-0.98

 well

-0.94

POSITIVE LOGITS

只是一个

1.21

still

1.13

 wciąż

1.12

Seems

1.09

 сих

1.06

 vẫn

1.05

 hiçbir

1.04

ほとん

1.02

 почему

1.01

 STILL

1.01

Activations Density 0.065%