INDEX

Explanations

constant speed and acceleration

The neuron activates on technical action terms and procedural words describing specific operations in patent-style method descriptions.

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

 avantage

-0.79

副作用

-0.79

Lean

-0.78

nika

-0.73

 adapter

-0.73

 courtroom

-0.73

 Lean

-0.71

 July

-0.71

Eleven

-0.71

July

-0.69

POSITIVE LOGITS

 acceleration

0.99

 dwells

0.98

acceleration

0.98

 ramp

0.96

 dwell

0.95

 Dwell

0.93

 Acceleration

0.93

Acceleration

0.90

 ramps

0.84

 starting

0.81

Activations Density 0.020%