INDEX

Explanations

I

np_max-act · gemini-2.0-flash

This neuron activates on the little floating‐point numbers and header markers (like “<|end_header_id|>” and the subsequent log‐prob scores) that appear right before each assistant reply.

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

questions and suggestions related to planning and preparation.

oai_token-act-pair · gpt-4o-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

 dài

-0.07

peek

-0.07

module

-0.06

こそ

-0.06

درس

-0.06

 Greenwood

-0.06

्वच

-0.06

QL

-0.06

appropriate

-0.06

processing

-0.06

POSITIVE LOGITS

 reactions

0.07

.neo

0.06

 imposition

0.06

 Trib

0.06

 труб

0.06

 searchBar

0.06

 Müslüman

0.06

Syn

0.06

CurrentValue

0.06

Neg

0.06

Activations Density 0.133%

I

This neuron activates on the little floating‐point numbers and header markers (like “<|end_header_id|>” and the subsequent log‐prob scores) that appear right before each assistant reply.

questions and suggestions related to planning and preparation.

No Comments

No Known Activations

I

This neuron activates on the little floating‐point numbers and header markers (like “<|end_header_id|>” and the subsequent log‐prob scores) that appear right before each assistant reply.

questions and suggestions related to planning and preparation.

No Comments

No Known Activations