INDEX

Explanations

47

np_max-act · gemini-2.0-flash

The neuron activates on the internal system‐ and header‐marking tokens that frame the conversation (e.g. <|start_header_id|>, <|end_header_id|>, <|eot_id|>, etc.).

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 timber

-0.07

rew

-0.06

 ());↵↵

-0.06

;p

-0.06

 sharpen

-0.06

	renderer

-0.06

 Serena

-0.06

 trail

-0.06

eview

-0.06

ariat

-0.06

POSITIVE LOGITS

 karşı

0.07

 ку

0.06

 kaum

0.06

 آن

0.06

_MM

0.06

 Мы

0.06

NEC

0.06

.qual

0.06

 στον

0.06

 letz

0.06

Activations Density 0.035%

47

The neuron activates on the internal system‐ and header‐marking tokens that frame the conversation (e.g. <|start_header_id|>, <|end_header_id|>, <|eot_id|>, etc.).

No Comments

No Known Activations