INDEX

Explanations

or

np_max-act · gemini-2.0-flash

This neuron activates on words used in the assistant’s refusal or policy-violation responses (e.g. “cannot,” “pretend,” “girlfriend,” “generate,” etc.).

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

requests for inappropriate or unethical content.

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 kromě

-0.06

 sinful

-0.06

dz

-0.06

 захисту

-0.06

 никто

-0.05

ossible

-0.05

іх

-0.05

_KHR

-0.05

.getDocument

-0.05

 Серед

-0.05

POSITIVE LOGITS

Sort

0.07

 printers

0.07

 اش

0.07

(cm

0.06

 وي

0.06

(Seq

0.06

(*.

0.06

_COMMON

0.06

;"><?

0.06

 drown

0.06

Activations Density 0.051%

or

This neuron activates on words used in the assistant’s refusal or policy-violation responses (e.g. “cannot,” “pretend,” “girlfriend,” “generate,” etc.).

requests for inappropriate or unethical content.

No Comments

No Known Activations