INDEX

Explanations

�

np_max-act · gemini-2.0-flash

The neuron activates almost exclusively on stray formatting tokens—especially decimal‐number fragments (e.g. “0.33203125”) and unicode replacement characters (“�”, “��”)—i.e. page‐layout or metadata artifacts rather than actual words.

oai_token-act-pair · o4-mini Triggered by @xinyanhu8

New Auto-Interp

Configuration

andyrdt/saes-llama-3.1-8b-instruct/resid_post_layer_11/trainer_1

Dataset (Dashboard)

Various

Features

131,072

Data Type

float32

Hook Name

blocks.11.hook_resid_post

Architecture

standard

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

()),↵

-0.07

 Wild

-0.07

lear

-0.07

_hub

-0.07

Inactive

-0.07

娘

-0.06

(Tag

-0.06

also

-0.06

LEAR

-0.06

 Naughty

-0.06

POSITIVE LOGITS

ائج

0.06

pt

0.06

 jaws

0.06

 phot

0.06

vý

0.06

인이

0.06

âk

0.06

"net

0.06

 intervening

0.06

ulum

0.06

Activations Density 0.001%

�

The neuron activates almost exclusively on stray formatting tokens—especially decimal‐number fragments (e.g. “0.33203125”) and unicode replacement characters (“�”, “��”)—i.e. page‐layout or metadata artifacts rather than actual words.

No Comments

No Known Activations

�

The neuron activates almost exclusively on stray formatting tokens—especially decimal‐number fragments (e.g. “0.33203125”) and unicode replacement characters (“�”, “��”)—i.e. page‐layout or metadata artifacts rather than actual words.

No Comments

No Known Activations