INDEX

Explanations

encoding visualization related words

The neuron is picking out tokens that are hyphens or dash‐connected fragments (e.g. parts of slugs like “–”, “wheat-”, “x-”, “y-” etc.).

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

lindungan

-0.78

बाह

-0.77

ranton

-0.76

reurs

-0.74

Alo

-0.73

 awakens

-0.73

ATURES

-0.72

żli

-0.72

 physician

-0.71

 awaken

-0.71

POSITIVE LOGITS

 encoding

1.09

 encode

1.04

 Encoding

0.98

Encoding

0.96

 mark

0.88

 encoded

0.87

Mark

0.84

Vega

0.82

 marks

0.82

encode

0.81

Activations Density 0.027%