INDEX

Explanations

questions or gratitude

np_max-act · gemini-2.0-flash

This neuron detects the end‐of‐user‐input marker (the <|eot_id|> token) indicating the close of a user’s turn.

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

sentences where the user asks the assistant for help or advice (direct first-person requests/questions).

oai_token-act-pair · gpt-5-mini Triggered by @vetterc0

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

'];?>↵

-0.07

.parameter

-0.06

.AspNet

-0.06

 Creature

-0.06

 correctly

-0.06

_rest

-0.06

"]

-0.06

.showToast

-0.06

 Continuous

-0.06

kan

-0.06

POSITIVE LOGITS

横

0.07

.FONT

0.06

jít

0.06

ihr

0.06

 Realt

0.06

جا

0.06

(view

0.06

 roky

0.06

 jednot

0.06

BCH

0.06

Activations Density 0.047%

questions or gratitude

This neuron detects the end‐of‐user‐input marker (the <|eot_id|> token) indicating the close of a user’s turn.

sentences where the user asks the assistant for help or advice (direct first-person requests/questions).

No Comments

No Known Activations