INDEX

Explanations

code

np_max-act · gemini-2.0-flash

This neuron activates on special “structural” tokens and role markers (e.g. <|start_header_id|>, <|end_header_id|>, “user”, “assistant”, <|eot_id|>, “Answer”), i.e. it detects conversation-turn or header delimiters rather than normal text.

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

 Buzz

-0.06

plets

-0.06

IW

-0.06

CARD

-0.06

.assertNotNull

-0.06

housing

-0.06

::::

-0.06

xyz

-0.06

 RFID

-0.05

 afterward

-0.05

POSITIVE LOGITS

 Investig

0.08

 ecological

0.08

aceutical

0.07

Orig

0.07

 certified

0.07

.Document

0.06

 Expr

0.06

nesota

0.06

MATRIX

0.06

prm

0.06

Activations Density 0.027%

code

This neuron activates on special “structural” tokens and role markers (e.g. <|start_header_id|>, <|end_header_id|>, “user”, “assistant”, <|eot_id|>, “Answer”), i.e. it detects conversation-turn or header delimiters rather than normal text.

No Comments

No Known Activations

code

This neuron activates on special “structural” tokens and role markers (e.g. <|start_header_id|>, <|end_header_id|>, “user”, “assistant”, <|eot_id|>, “Answer”), i.e. it detects conversation-turn or header delimiters rather than normal text.

No Comments

No Known Activations