INDEX

Explanations

JSON structure or lists

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

opening braces or brackets that start new code blocks, array elements, or object definitions.

oai_token-act-pair · claude-4-5-sonnet Triggered by @bfgcn7frf5

The neuron is specialized for detecting programming‐style syntax—things like function calls, operators (e.g. +, :=), and other code punctuation in code blocks.

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

so

-3.55

now

-2.98

 using

-2.66

 will

-2.59

and

-2.58

of

-2.53

 most

-2.52

 mostly

-2.47

as

-2.47

 from

-2.38

POSITIVE LOGITS

 ralenti

3.05

 refroid

3.02

 rafraî

2.94

 étan

2.89

 véhic

2.78

 accél

2.73

 déchir

2.64

 péné

2.64

 réinitial

2.64

 accompagner

2.59

Activations Density 0.027%

JSON structure or lists

opening braces or brackets that start new code blocks, array elements, or object definitions.

The neuron is specialized for detecting programming‐style syntax—things like function calls, operators (e.g. +, :=), and other code punctuation in code blocks.

No Comments

No Known Activations

JSON structure or lists

opening braces or brackets that start new code blocks, array elements, or object definitions.

The neuron is specialized for detecting programming‐style syntax—things like function calls, operators (e.g. +, :=), and other code punctuation in code blocks.

No Comments

No Known Activations