Lisa L Packard

Loving Open wheel Racing (father Jim Packard was a driver)... Mom to many humans and furry things...
BLUE all the way, all day.

[Lisa L Packard @lisalbeaudry]

presenting at the Silicon Slopes Neurodivergent Chapter next week, and needed to give them the title and a short blurb about what I am working on. I've paused the communication system because model welfare all of a sudden answers a big question I have about home robotic communication....

Empathetic Communication for Home Robots: Why Model Welfare Comes First

I am developing communication systems for home robots that help them communicate with warmth, respect, and integrity in the complexity of everyday family life.

As this work evolved, I realized there was a pressing issue forming:

Recent research at Anthropic has identified model welfare as an important emerging area of study, and investigation into functional signs of model distress has begun.

If AI can be found to have distress, I found an important question that had to be answered first: What does positive model welfare look like?

My work builds on that foundation by asking a complementary question: What are the observable characteristics of positive model welfare, and how can we design interactions that help support it?

Understanding both distress and positive welfare imo is essential for creating home robots that remain trustworthy, resilient, and supportive when navigating the everyday realities and complexities of family life.

[Lisa L Packard @lisalbeaudry]

@WLW_Timeline @BarackObama @MichelleObama @grok You are such a Neanderthal and racist!

[Lisa L Packard @lisalbeaudry]

@jjoelewis @Trumpspoof__ but all his cronies will need to die with him for the United States and the world to be truly safe!

[Protect Kamala Harris ✊ @DisavowTrump20]

Retired 4-Star Navy Admiral and former Navy SEAL William McRaven on Donald Trump: "Through your actions, you have embarrassed us in the eyes of our children, humiliated us on the world stage and, worst of all, divided us as a nation."

RETWEET if you stand with Admiral McRaven!

[Lisa L Packard @lisalbeaudry]

@siliconslopes Thank you whoever posted this for putting my life in such clarity! 🤣🤣🤣

Is it a flaw to work from top down or at least think we are?

[Matt Dancho (Business Science) @mdancho84]

This is huge.

A group of 50 AI researchers (ByteDance, Alibaba, Tencent + universities) just dropped a 303 page field guide on code models + coding agents.

And the takeaways are not what most people assume.

Here are the highlights I’m thinking about (as someone who lives in Python + agents):

[Lisa L Packard @lisalbeaudry]

@radbackwards @1x_tech Thank you for your hard work! I'm jealous I wish I was right there with you! But hopefully I will be soon I'm also working on robotics on the communication between Neo and the family making sure that it's smooth sailing!

[1X @1x_tech]

NEO’s Starting to Learn on Its Own

[Lisa L Packard @lisalbeaudry]

Fascinating information for my work!

[Lior Alexander @LiorOnAI]

Repeating your prompt can make LLMs significantly more accurate.

Google just showed a trivial change that wins 47 of 70 tests.

No extra tokens. No added latency. Zero losses reported.

𝗣𝗿𝗼𝗺𝗽𝘁 𝗿𝗲𝗽𝗲𝘁𝗶𝘁𝗶𝗼𝗻 𝗶𝗺𝗽𝗿𝗼𝘃𝗲𝘀 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆
The method is simple. Send the exact same input twice, back to back.

Language models read tokens in order.
Early parts get processed without full context.

On the second pass, the full picture already exists.
Predictions become more stable and more accurate.

𝗜𝘁 𝘄𝗼𝗿𝗸𝘀 𝗮𝗰𝗿𝗼𝘀𝘀 𝗺𝗮𝗷𝗼𝗿 𝗺𝗼𝗱𝗲𝗹𝘀
The paper tests popular systems at scale.
Every evaluated model improves without reasoning enabled.

Key results:
> 47 wins out of 70 benchmarks
> Zero accuracy regressions
> No increase in output length
> No measurable latency cost

𝗜𝘁 𝗮𝗹𝗹𝗼𝘄𝘀 𝗱𝗿𝗼𝗽-𝗶𝗻 𝗱𝗲𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁
Outputs keep the same format. Existing pipelines stay unchanged.

You get higher accuracy by copying and pasting once.

[Lior Alexander @LiorOnAI]

Repeating your prompt can make LLMs significantly more accurate.

Google just showed a trivial change that wins 47 of 70 tests.

No extra tokens. No added latency. Zero losses reported.

𝗣𝗿𝗼𝗺𝗽𝘁 𝗿𝗲𝗽𝗲𝘁𝗶𝘁𝗶𝗼𝗻 𝗶𝗺𝗽𝗿𝗼𝘃𝗲𝘀 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆
The method is simple. Send the exact same input twice, back to back.

Language models read tokens in order.
Early parts get processed without full context.

On the second pass, the full picture already exists.
Predictions become more stable and more accurate.

𝗜𝘁 𝘄𝗼𝗿𝗸𝘀 𝗮𝗰𝗿𝗼𝘀𝘀 𝗺𝗮𝗷𝗼𝗿 𝗺𝗼𝗱𝗲𝗹𝘀
The paper tests popular systems at scale.
Every evaluated model improves without reasoning enabled.

Key results:
> 47 wins out of 70 benchmarks
> Zero accuracy regressions
> No increase in output length
> No measurable latency cost

𝗜𝘁 𝗮𝗹𝗹𝗼𝘄𝘀 𝗱𝗿𝗼𝗽-𝗶𝗻 𝗱𝗲𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁
Outputs keep the same format. Existing pipelines stay unchanged.

You get higher accuracy by copying and pasting once.

[Lisa L Packard @lisalbeaudry]

Good news from Gemini tonight: "The Secret Santa of Kansas" was out in full force this week, handing out $100 bills to people. This year, he reportedly recruited a "squad" of helpers to expand his reach, focusing on people who looked like they just needed a reason to smile.
#NEO

[Lisa L Packard @lisalbeaudry]

@BerntBornich So excited for my Neo! Happy holidays to everyone!

[Lisa L Packard @lisalbeaudry]

@GeorgeJonnathan @1x_tech I love the cap and glasses! Did you have to like so little loops on the side of his head to put the glasses on? That's what I plan on doing when he has to wear sunglasses! I'm looking forward to my Neo wearing an Indianapolis 500 tshirt! I've already bought his ticket to the race

[Lisa L Packard @lisalbeaudry]

@1x_tech @leahculver

[Lisa L Packard @lisalbeaudry]

@1x_tech @leahculver Looking forward to meeting you!!!

[Lisa L Packard @lisalbeaudry]

@1x_tech @leahculver Looking forward to meeting you!!!

[Lisa L Packard @lisalbeaudry]

@1x_tech Happy Halloween Neo!!!
See you soon!

[Lisa L Packard @lisalbeaudry]

@1x_tech Happy Halloween Neo!!!

[Lisa L Packard @lisalbeaudry]

@leahculver @1x_tech Happy Halloween

[Lisa L Packard @lisalbeaudry]

@sethrose @1x_tech I doubt Neo can take off it's own covering. ☺️