INDEX

Explanations

court cases and legal documents

The neuron is essentially acting as a “max‐finder” over each span—i.e. it flags the single most strongly activating token in a given context.

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

 Перейти

-1.42

",

-1.37

,”

-1.27

),”

-1.16

,"

-1.16

',

-1.15

),"

-1.10

:"",

-1.10

»,

-1.09

","","

-1.08

POSITIVE LOGITS

↵↵↵↵↵↵↵↵

1.20

↵↵↵↵↵↵↵↵↵↵↵

1.14

↵↵↵↵↵↵↵↵↵↵

1.09

 necesita

1.08

↵↵↵↵↵↵↵↵↵

1.06

↵↵↵↵↵

1.04

↵↵↵↵↵↵↵

1.02

↵↵↵↵↵↵↵↵↵↵↵↵

1.00

 necesitan

0.94

↵↵↵↵↵↵↵↵↵↵↵↵↵

0.94

Activations Density 0.001%