INDEX

Explanations

Testing and code

np_max-act · gemini-2.0-flash

This neuron activates on the special “agent‐style” metadata labels (e.g., Thought, Action, Observation, Final Answer, etc.) used to structure the step-by-step reasoning format.

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

 sentencing

-0.07

 Learning

-0.07

 тоб

-0.06

_waiting

-0.06

 Brow

-0.06

dete

-0.06

 نبود

-0.06

 Arbor

-0.06

ora

-0.06

 пла

-0.06

POSITIVE LOGITS

另外

0.06

 Eventually

0.06

 )↵↵↵

0.06

 ofrece

0.06

.bank

0.06

 збільш

0.06

-,

0.06

_bootstrap

0.06

θυν

0.05

 generated

0.05

Activations Density 0.004%

Testing and code

This neuron activates on the special “agent‐style” metadata labels (e.g., Thought, Action, Observation, Final Answer, etc.) used to structure the step-by-step reasoning format.

No Comments

No Known Activations

Testing and code

This neuron activates on the special “agent‐style” metadata labels (e.g., Thought, Action, Observation, Final Answer, etc.) used to structure the step-by-step reasoning format.

No Comments

No Known Activations