INDEX

Explanations

Java code

np_max-act · gemini-2.0-flash

This neuron primarily responds to very short, uppercase (or partly uppercase) identifier tokens—e.g. single‐letter or two‐letter code acronyms.

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

_FACTOR

-0.07

setValue

-0.07

Compar

-0.07

protobuf

-0.07

 ánh

-0.06

.member

-0.06

jších

-0.06

_FIND

-0.06

 Mitar

-0.06

 bahis

-0.06

POSITIVE LOGITS

ภ

0.07

تیجه

0.06

ステ

0.06

_ST

0.06

 efect

0.06

.Results

0.06

}`);↵

0.06

 stát

0.06

tempts

0.06

 mezi

0.06

Activations Density 0.013%

Java code

This neuron primarily responds to very short, uppercase (or partly uppercase) identifier tokens—e.g. single‐letter or two‐letter code acronyms.

No Comments

No Known Activations