INDEX

Explanations

clinical predictors

np_acts-logits-general · gemini-2.5-flash-lite

line numbers or numeric references in code diffs, legal documents, and technical files.

oai_token-act-pair · claude-3-7-sonnet-20250219 Triggered by @neilrathi

The neuron detects the numeric line‐range or hunk‐size values in unified “diff” headers (e.g. the “38” in “@@ -1,3 +1,38 @@”).

oai_token-act-pair · o4-mini Triggered by @jyhe0408

New Auto-Interp

Configuration

google/gemma-scope-27b-pt-res/layer_10/width_131k

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 offrir

-1.64

umerable

-1.63

nô

-1.62

 interesar

-1.57

 offenders

-1.57

绸

-1.55

都没

-1.55

และ

-1.55

니메이션

-1.54

からない

-1.53

POSITIVE LOGITS

2.00

has

1.86

we

1.65

 становится

1.59

 used

1.57

1.55

 continues

1.55

1.52

ting

1.50

For

1.50

Activations Density 0.000%

clinical predictors

line numbers or numeric references in code diffs, legal documents, and technical files.

The neuron detects the numeric line‐range or hunk‐size values in unified “diff” headers (e.g. the “38” in “@@ -1,3 +1,38 @@”).

No Comments

No Known Activations

clinical predictors

line numbers or numeric references in code diffs, legal documents, and technical files.

The neuron detects the numeric line‐range or hunk‐size values in unified “diff” headers (e.g. the “38” in “@@ -1,3 +1,38 @@”).

No Comments

No Known Activations