INDEX

Explanations

Attacks and danger

np_max-act · gemini-2.0-flash

The neuron selectively activates on header‐ and metadata‐style tokens and delimiters (e.g. <|start_header_id|>, “Ways,” “in,” “the,” “Zone”), indicating it’s tuned to detect document headers or structured metadata markers.

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

ه

-0.08

fil

-0.07

 illustrate

-0.06

ujícím

-0.06

xEE

-0.06

 chat

-0.06

ього

-0.05

uit

-0.05

ْد

-0.05

 Trust

-0.05

POSITIVE LOGITS

>↵↵↵

0.07

 Hayward

0.07

_register

0.07

OSH

0.07

 tipo

0.07

.tolist

0.06

"}>↵

0.06

く

0.06

�

0.06

iş

0.06

Activations Density 0.016%

Attacks and danger

The neuron selectively activates on header‐ and metadata‐style tokens and delimiters (e.g. <|start_header_id|>, “Ways,” “in,” “the,” “Zone”), indicating it’s tuned to detect document headers or structured metadata markers.

No Comments

No Known Activations

Attacks and danger

The neuron selectively activates on header‐ and metadata‐style tokens and delimiters (e.g. <|start_header_id|>, “Ways,” “in,” “the,” “Zone”), indicating it’s tuned to detect document headers or structured metadata markers.

No Comments

No Known Activations