INDEX

Explanations

yyyy

np_max-act · gemini-2.0-flash

The neuron activates on the special role‐ and metadata‐marker tokens (e.g. the `<|start_header_id|>`, `<|end_header_id|>`, “system,” “user,” and similar control tokens) rather than on normal text content.

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

-rule

-0.07

 подоб

-0.07

플

-0.06

 gắng

-0.06

chart

-0.06

 seeks

-0.06

ystems

-0.06

markers

-0.06

POSITIVE LOGITS

yyyy

0.07

 yyyy

0.06

 Armenian

0.06

agne

0.06

voy

0.06

iP

0.06

 Relay

0.06

 synchronous

0.06

 adec

0.06

PAY

0.06

Activations Density 0.003%

yyyy

The neuron activates on the special role‐ and metadata‐marker tokens (e.g. the `<|start_header_id|>`, `<|end_header_id|>`, “system,” “user,” and similar control tokens) rather than on normal text content.

No Comments

No Known Activations