Connect with us

Technology

AutoToS makes LLM planning fast, accurate and inexpensive

Published

on

AutoToS makes LLM planning fast, accurate and inexpensive

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More


Large language models (LLMs) have shown promise in solving planning and reasoning tasks by searching through possible solutions. However, existing methods can be slow, computationally expensive and provide unreliable answers. 

Researchers from Cornell University and IBM Research have introduced AutoToS, a new technique that combines the planning power of LLMs with the speed and accuracy of rule-based search algorithms. AutoToS eliminates the need for human intervention and significantly reduces the computational cost of solving planning problems. This makes it a promising technique for LLM applications that must reason over large solution spaces.

There is a growing interest in using LLMs to handle planning problems, and researchers have developed several techniques for this purpose. The more successful techniques, such as Tree of Thoughts, use LLMs as a search algorithm that can validate solutions and propose corrections.

Advertisement

While these approaches have demonstrated impressive results, they face two main challenges. First, they require numerous calls to LLMs, which can be computationally expensive, especially when dealing with complex problems with thousands of possible solutions. Second, they do not guarantee that the LLM-based algorithm qualifies for “completeness” and “soundness.” Completeness ensures that if a solution exists, the algorithm will eventually find it, while soundness guarantees that any solution returned by the algorithm is valid.

Thought of Search (ToS) offers an alternative approach. ToS leverages LLMs to generate code for two key components of search algorithms: the successor function and the goal function. The successor function determines how the search algorithm explores different nodes in the search space, while the goal function checks whether the search algorithm has reached the desired state. These functions can then be used by any offline search algorithm to solve the problem. This approach is much more efficient than keeping the LLM in the loop during the search process.

“Historically, in the planning community, these search components were either manually coded for each new problem or produced automatically via translation from a description in a planning language such as PDDL, which in turn was either manually coded or learned from data,” Michael Katz, principal research staff member at IBM Research, told VentureBeat. “We proposed to use the large language models to generate the code for the search components from the textual description of the planning problem.”

The original ToS technique showed impressive progress in addressing the soundness and completeness requirements of search algorithms. However, it required a human expert to provide feedback on the generated code and help the model refine its output. This manual review was a bottleneck that reduced the speed of the algorithm.

Advertisement

Automating ToS

AutoToS
AutoToS (source: arXiv)

“In [ToS], we assumed a human expert in the loop, who could check the code and feedback the model on possible issues with the generated code, to produce a better version of the search components,” Katz said. “We felt that in order to automate the process of solving the planning problems provided in a natural language, the first step must be to take the human out of that loop.”

AutoToS automates the feedback and exception handling process using unit tests and debugging statements, combined with few-shot and chain-of-thought (CoT) prompting techniques.

AutoToS works in multiple steps. First, it provides the LLM with the problem description and prompts it to generate code for the successor and goal functions. Next, it runs unit tests on the goal function and provides feedback to the model if it fails. The model then uses this feedback to correct its code. Once the goal function passes the tests, the algorithm runs a limited breadth-first search to check if the functions are sound and complete. This process is repeated until the generated functions pass all the tests. 

Finally, the validated functions are plugged into a classic search algorithm to perform the full search efficiently.

AutoToS in action

The researchers evaluated AutoToS on several planning and reasoning tasks, including BlocksWorld, Mini Crossword and 24 Game. The 24 Game is a mathematical puzzle where you are given four integers and must use basic arithmetic operations to create a formula that equates to 24. BlocksWorld is a classic AI planning domain where the goal is to rearrange blocks stacked in towers. Mini Crosswords is a simplified crossword puzzle with a 5×5 grid.

Advertisement

They tested various LLMs from different families, including GPT-4o, Llama 2 and DeepSeek Coder. They used both the largest and smallest models from each family to evaluate the impact of model size on performance.

Their findings showed that with AutoToS, all models were able to identify and correct errors in their code when given feedback. The larger models generally produced correct goal functions without feedback and required only a few iterations to refine the successor function. Interestingly, GPT-4o-mini performed surprisingly well in terms of accuracy despite its small size.

“With just a few calls to the language model, we demonstrate that we can obtain the search components without any direct human-in-the-loop feedback, ensuring soundness, completeness, accuracy and nearly 100% accuracy across all models and all domains,” the researchers write.

Compared to other LLM-based planning approaches, ToS drastically reduces the number of calls to the LLM. For example, for the 24 Game dataset, which contains 1,362 puzzles, the previous approach would call GPT-4 approximately 100,000 times. AutoToS, on the other hand, needed only 2.2 calls on average to generate sound search components.

Advertisement

“With these components, we can use the standard BFS algorithm to solve all the 1,362 games together in under 2 seconds and get 100% accuracy, neither of which is achievable by the previous approaches,” Katz said.

AutoToS for enterprise applications

AutoToS can have direct implications for enterprise applications that require planning-based solutions. It cuts the cost of using LLMs and reduces the reliance on manual labor, enabling experts to focus on high-level planning and goal specification.

“We hope that AutoToS can help with both the development and deployment of planning-based solutions,” Katz said. “It uses the language models where needed—to come up with verifiable search components, speeding up the development process and bypassing the unnecessary involvement of these models in the deployment, avoiding the many issues with deploying large language models.”

ToS and AutoToS are examples of neuro-symbolic AI, a hybrid approach that combines the strengths of deep learning and rule-based systems to tackle complex problems. Neuro-symbolic AI is gaining traction as a promising direction for addressing some of the limitations of current AI systems.

Advertisement

“I don’t think that there is any doubt about the role of hybrid systems in the future of AI,” Harsha Kokel, research scientist at IBM, told VentureBeat. “The current language models can be viewed as hybrid systems since they perform a search to obtain the next tokens.”

While ToS and AutoToS show great promise, there is still room for further exploration.

“It is exciting to see how the landscape of planning in natural language evolves and how LLMs improve the integration of planning tools in decision-making workflows, opening up opportunities for intelligent agents of the future,” Kokel and Katz said. “We are interested in general questions of how the world knowledge of LLMs can help improve planning and acting in real-world environments.”


Source link
Continue Reading
Advertisement
Click to comment

You must be logged in to post a comment Login

Leave a Reply

Servers computers

Dell EMC Storage Expansion Enclosure ME424 Rack Server rack server 42u dell case cabinet

Published

on

Dell EMC Storage Expansion Enclosure  ME424 Rack Server rack server 42u dell case cabinet



Quality Rack Server from China.
https://www.opticaltransceiver-module.com/china-dell_emc_storage_expansion_enclosure_me424_rack_server_rack_server_42u_dell_case_cabinet-37741961.html
LonRise Equipment Co. Ltd. quality manufacturer from China.
We can supply:
Optical Transceiver Module : https://www.opticaltransceiver-module.com/supplier-optical_transceiver_module-117281.html
SFP Optical Transceiver : https://www.opticaltransceiver-module.com/supplier-sfp_optical_transceiver-117295.html
Cisco SFP Modules : https://www.opticaltransceiver-module.com/supplier-cisco_sfp_modules-117282.html
Welcome to visit our official website : http://www.opticaltransceiver-module.com .

source

Continue Reading

Technology

Smart TVs take snapshots of what you watch multiple times per second

Published

on

Smart TVs take snapshots of what you watch multiple times per second

Smart TVs watch everything on the screen

Shutterstock

Popular smart TV models made by Samsung and LG can take multiple snapshots of what you are watching every second – even when they are being used as external displays for your laptop or video game console.

Smart TV manufacturers use these frequent screenshots, as well as audio recordings, in their automatic content recognition systems, which track viewing habits in order to target people with specific advertising. But researchers showed this tracking by some of the world’s most popular smart TV brands – Samsung TVs can take screenshots every 500 milliseconds and LG…

Advertisement

Source link

Continue Reading

Technology

Astronaut’s stunning photo shows ‘flowing silver snakes’

Published

on

Astronaut's stunning photo shows 'flowing silver snakes'

Over his three previous missions to the International Space Station (ISS), NASA astronaut Don Pettit earned a reputation for having a keen eye when it comes to photographing Earth and beyond.

Since arriving at the ISS on his fourth orbital mission earlier this month, Pettit, who at 69 is NASA’s oldest active astronaut, has wasted little time in grabbing the station’s cameras to capture and share fresh dazzling imagery shot from 250 miles above Earth.

In his latest work posted on X (formerly Twitter) on Tuesday, Pettit shared some sublime and highly artistic shots showing moonlight reflecting off of locations in the Amazon basin in South America.

“Somewhere over the Amazon basin, shooting photos of cities at night, I noticed the light from a near-full moon reflecting off of the meandering rainforest rivers,” Pettit wrote in the post, describing how the waterwaays appeared as “flowing silver snakes” and “glowing golden claws.”

Moonshine from space. Somewhere over the Amazon basin, shooting photos of cities at night, I noticed the light from a near-full moon reflecting off of the meandering rainforest rivers. In the cool moon-ish light these rivers became flowing silver snakes. When the moonlight was… pic.twitter.com/SGIUAJLhpP

— Don Pettit (@astro_Pettit) September 24, 2024

Advertisement

As usual, Pettit shared details of the equipment and camera settings that he used to grab the shots. In this case, he used a full-frame Nikon Z9 body with a 200mm lens set at f/2 and a speed of 1/320 sec, while the ISO was set at 25600. Pettit added that he processed the images in Photoshop before sharing them.

While Pettit is also working on science research aboard the ISS along with all of the other astronauts there, he also loves to head to the station’s seven-window Cupola module to capture extraordinary views of Earth using the Nikon Z9. Just recently he shared a striking shot of London at night, and, in another remarkable image, managed to capture the Polaris Dawn Crew Dragon capsule as it entered Earth’s atmosphere at high speed at the end of a historic five-day mission.

Pettit’s will be in orbit until March 2025 — ample time to create more works of art from space.






Source link

Advertisement

Continue Reading

Servers computers

Mini Rack, HomeLab Stack – Mini Server Rack

Published

on

Mini Rack, HomeLab Stack - Mini Server Rack



Let’s try this again, but WITHOUT the sponsorship. There’s building a MINI SERVER RACK and then there’s beating Raid Owl in the mini server rack build challenge. Let’s see if I can do both.

📦 Mini Server Rack Parts List 📦
– DeskPi RackMate T1: https://amzn.to/3zVZk9w
– DeskPi Rackmate Shelves: https://amzn.to/4di9gIZ
– Protectli VP1210: https://amzn.to/4bWEyUF
– Beelink EQ13: https://amzn.to/3Wfu9gR
– RealHD 2.5G Switch: https://amzn.to/46gA3TG
– Tenda 2.5G Switch (won’t fit in rack but dual SFP+): https://amzn.to/3SlShx4
– PiKVM: https://pikvm.org/
– Axe Effect by Craft Computing: https://www.craftcomputing.com/product/axe-effect-temperature-sensor-beta-/1?cs=true&cst=custom
– EZCOO KVM: https://amzn.to/3WgjO4u
– 120mm USB Ran: https://amzn.to/3YhsIRL
– Noctua 5V 120mm fan: https://amzn.to/46kFS2c
– Rii X8 Mini Keyboard: https://amzn.to/3WkQmtY
– USB to SATA adapter: https://amzn.to/46gWWGu
– 12″ HDMI Cables: https://amzn.to/3LBT3T3
– SmallRig Monitor Clamp (Magic Arm): https://amzn.to/3Skj9hh
– Cable Organizer Box: https://amzn.to/3ylG4BP
– Cable clips: https://amzn.to/3WkFYSW
– 280W USB GaN III Charger: https://amzn.to/3LzF6Fl
– Anker Surge Protector: https://amzn.to/3LzFvaP

(Affiliate links may be included in this description. I may receive a small commission at no cost to you.)

Check out Raid Owl’s build here: https://www.youtube.com/watch?v=wJUDhQ7s9HM

Support me on Patreon: https://www.patreon.com/technotim
Sponsor me on GitHub: https://github.com/sponsors/timothystewart6
Subscribe on Twitch: https://www.twitch.tv/technotim
Become a YouTube member: https://www.youtube.com/channel/UCOk-gHyjcWZNj3Br4oxwh0A/join
Merch Shop 🛍️: https://l.technotim.live/shop
Gear Recommendations: https://l.technotim.live/gear
Get Help in Our Discord Community: https://l.technotim.live/discord
2nd channel: https://www.youtube.com/@TechnoTimTinkers

00:00 – Intro
01:34 – Mini Server Rack
02:48 – Network Firewall
04:24 – Cooling
04:56 – Networking
05:40 – Compute & Clients
07:35 – Servers
08:44 – Mini ITX Cluster ARM Board
10:01 – Retro Gaming & More
11:17 – DietPi & PiOS
11:31 – 4 Port KVM
11:49 – KVM over IP
13:53 – Dual Monitors (and Dashboards)
15:23 – Monitor Arm
15:45 – Temperature Monitoring
16:42 – My Thoughts & Regrets

Thank you for watching! .

source

Continue Reading

Technology

Apple iPhone 16 vs Apple iPhone 15

Published

on

Apple iPhone 16 vs Apple iPhone 15

Apple launched four smartphones earlier this month. The iPhone 16 is the most affordable one out of the bunch, the entry-level model. In this article, we’ll compare it to its predecessor, so we’ll compare the Apple iPhone 16 vs Apple iPhone 15. Not much has changed compared to last year’s model, but there are some differences worth talking about, though, of course, hence the comparison.

Some of you may even wonder if it’s worth getting last year’s model now, instead of going for a new one. Well, we do hope that you’ll find this comparison useful. That it will help you make that sort of decisions, as going for the latest model is not always the best idea. We will first list the specifications of both smartphones and will then compare them across a number of other categories. With that being said, let’s get down to it.

Specs

Apple iPhone 16 vs Apple iPhone 15, respectively

Screen size:
6.1-inch Super Retina XDR OLED ( flat, 60Hz, HDR, 2,000 nits)
6.1-inch Super Retina XDR display (60Hz, HDR, 2,000 nits)
Display resolution:
2556 x 1179
2556 x 1179
SoC:
Apple A18 (3nm)
Apple A16 Bionic
RAM:
8GB
6GB
Storage:
128GB/256GB/512GB (NVMe)
128GB/256GB/512GB
Rear cameras:
48MP (wide, f/1.6 aperture, 1/1.56-inch sensor, 1.0um pixel size, sensor-shift OIS), 12MP (ultrawide, f/2.2 aperture, 120-degree FoV, 0.7um pixel size, PDAF)
48MP (wide, f/1.6 aperture, 1/1.56-inch sensor, 1.0um pixel size, sensor-shift OIS), 12MP (ultrawide, 120-degree FoV, 0.7um pixel size, f/2.4 aperture)
Front cameras:
12MP (f/1.9 aperture, PDAF, 1/3.6-inch sensor size)
Battery:
3,561mAh
3,349mAh
Charging:
38W wired, 25W MagSafe wireless, 15W Qi2 wireless, 7.5W Qi wireless & 4.5W reverse wired charging
20W wired, 15W wireless, 4.5W reverse wired charging (charger not included)
Dimensions:
147.6 x 71.6 x 7.8 mm
147.6 x 71.6 x 7.8mm
Weight:
170 grams
171 grams
Connectivity:
5G, LTE, NFC, Wi-Fi, USB Type-C, Bluetooth 5.3
Security:
Face ID (3D facial scanning)
OS:
iOS 18
iOS 17
Price:
$799+
$799+
Buy:
Apple iPhone 16 (Apple)
iPhone 15 (Apple)

Advertisement

Apple iPhone 16 vs Apple iPhone 15: Design

These two smartphones do look very similar, but it’s also easy to tell them apart. How? Well, because of their rear cameras. We’ll get to that part shortly, let’s talk about the build first. Both phones are made out of aluminum and glass. Both of them have the same exact shape. Their sides are flat all around, but also rounded towards the edges so that they do not cut into your hand when you hold them.

The front and back sides are flat on both smartphones. Both of them have flat displays and a pill-shaped cutout on their displays. That cutout is called the ‘Dynamic Island’. The bezels around the display are uniform, and they’re thicker than on the ‘Pro’ models, but still quite thin. The iPhone 16 does have an extra button on the right side, the so-called Camera Control button. In addition to that, there’s the power key there, just as on the iPhone 15. On the left, both smartphones have the volume rocker buttons and the Action Button.

When we flip them over, we see the main difference in their designs. The iPhone 16 has two vertically aligned cameras, while the iPhone 15 has a diagonal setup. The main camera is the same on both, while there is a slight change with the ultrawide unit. Both camera islands sit in the top-left corner of the back. The two phones have the exact same dimensions, while the iPhone 16 is 1 gram lighter, so they’re basically the same in that regard too. Both smartphones are IP68 certified for water and dust resistance.

Apple iPhone 16 vs Apple iPhone 15: Display

These two smartphones have the exact same displays. You’ll find a 6.1-inch Super Retina XDR OLED display on both phones. That display does support HDR10 content, and Dolby Vision too. It goes up to 2,000 nits of peak brightness. This panel is flat, and it has a resolution of 2556 x 1179 pixels. The display aspect ratio is 19.5:9, while the screen-to-body ratio is at around 86% on both phones. The Ceramic Shield glass protects both displays, but a newer version is included on the iPhone 16.

Advertisement

iphone 16 finish select 202409 6 7inch teal

There is no visible difference between the two panels. That’s not surprising at all as they’re exactly the same. Both panels are quite sharp and have good viewing angles. They’re also vivid, and get bright enough, even outdoors. They do not offer high-frequency PWM dimming, though, if that’s something you need. They also have one major limitation that you may care about, even though most regular users do not… both displays are limited to a 60Hz refresh rate.

That was one of the main complaints on these two panels. For the price tag Apple is asking for them, you’d expect 120Hz displays at this point in time. But no, Apple has limited both smartphones to 60Hz panels. Most people don’t really care, nor do they notice the difference, so… it’s not surprising Apple went down this road again. Both displays are very good, though, despite that limitation.

Apple iPhone 16 vs Apple iPhone 15: Performance

The Apple A18 SoC fuels the iPhone 16. That is Apple’s new 3nm processor. It is coupled with 8GB of RAM and NVMe flash storage. The iPhone 15, on the other hand, is fueled by the Apple A16 Bionic chip, a 4nm processor. That phone also comes with 6GB of RAM and NVMe flash storage. Do note that the storage is not expandable on either smartphone, which is per course these days.

Both of these processors have a 5-core graphics card, though the Apple A18 is the more powerful chip. Still, both of these smartphones offer great performance in day-to-day use. They fly through everything you throw at them, and you’d be hard-pressed to notice the SoC difference during daily use. They can both handle whatever you throw at them. The iPhone 16 may load some apps a bit faster, but other than that, they’re on the same level in terms of general performance.

Advertisement

Something similar can be said for games. The iPhone 16 may get there a bit faster, but in terms of general gaming performance, both smartphones do a fantastic job. That goes for even the most demanding titles you’d find in the App Store. The Apple A16 Bionic is still a very powerful processor, so that’s not surprising at all. Yes, they both do get warm during gaming, but not too hot to hold, nor does that affect the performance.

Apple iPhone 16 vs Apple iPhone 15: Battery

The iPhone 16 features a 3,561mAh battery, while its predecessor comes with a 3,349mAh battery. This bump is very nice to see despite the fact the iPhone 16 does not have a larger display or anything of the sort. From what we’ve seen thus far, the iPhone 16 does offer better battery life. The difference is not that huge, but it sure is noticeable, which is great to hear, as the iPhone 15 did have the worst battery life out of all the iPhone 15 models. It was not spectacular by any means.

Getting through a day on a single charge with the iPhone 16 should be doable for the vast majority of users. Yes, you can kill this phone’s battery in a day, if you want, but it’s nowhere as easy to do as it was the before. Not only is the battery capacity higher but the iPhone 16 is more power efficient in general, so that’s not surprising.

Apple also improved the charging speed on the device. The iPhone 16 now supports up to 38W wired and 25W wireless (MagSafe) charging. 15W Qi2 and 7.5W Qi wireless charging is also supported, as is 4.5W reverse wired charging. The iPhone 15 is limited to 20W wired, 15W MagSafe and Qi2 wireless, 7.5W Qi charging, and 4.5W reverse wired charging. Neither of these phones comes with a charger in the box.

Advertisement

Apple iPhone 16 vs Apple iPhone 15: Cameras

The main camera is the same on both phones, while there is a slight change with the ultrawide unit. A 48-megapixel main camera with an f/1.6 aperture and a 1/1.56-inch sensor size sits in both phones. A 26mm lens is also included, and the pixel size is 1.0um. Dual Pixel PDAF is also a part of the package, as is sensor-shift OIS. The ultrawide camera, on the other hand, is a 12-megapixel unit on both phones with a 120-degree FoV and a 0.7um pixel size. A different aperture is used on the new model, and the new unit also supports macro photography via that camera.

iphone 16 2

You’ll be hard-pressed to see the difference in camera performance between the two. Well, you can if you shoot macro photos, for sure, but other than that, not so much. Both smartphones provide nice-looking photos that lean towards warmer color temperatures. Apple simply loves that yellow tint on photos. The HDR performance is good, but both phones have a tendency to put a lot of brightness in darker spots, which makes the photos look a bit flatter than they should have, even in daytime. They’re not exactly contrasty.

The low light performance is good on both phones. They both tend to brighten up such scenes quite a bit. The ultrawide photos are slightly different, though that difference is visible in lower light only. The iPhone 16 does have the edge, though as I said, there’s barely any difference unless you’re shooting macro photos. The video performance is identical between the two devices, and the same goes for selfies.

Audio

There are stereo speakers included on both of these smartphones. Those speakers are good, but not great. They are not amongst the loudest out there, but they’ll be plenty loud for most people. The sound output is also good but nothing to write home about. The speakers do have very similar output, it’s even possible Apple used identical ones in these two devices.

Advertisement

What about an audio jack? Well, there is none. Both devices do include a Type-C port at the bottom, though. So you can hook up your wired headphones that way if you have a dongle or Type-C headphones. If not, there’s always Bluetooth 5.3 which is included on both devices.

Source link

Continue Reading

Servers computers

I built a big Server Rack for my Home Lab!

Published

on

I built a big Server Rack for my Home Lab!



Learn more about Home Lab and Servers, and how you can start your own tech projects! This is the first time I’ve put a lot of time and effort into building a Server Rack for my Homelab! I talk about the different components and devices I’ve put into it. Doesn’t need to be a big server rack though, just start with a single server, like I did. #ServerRack #HomeServer #HomeLab

Kit Page (My Equipment): https://kit.co/thedigitallife

Teleport-*: http://goteleport.com/thedigitallife

Follow me:

TWITTER: https://twitter.com/christianlempa
INSTAGRAM: https://instagram.com/christianlempa
DISCORD: https://discord.gg/christian-lempa-s-tech-community-702179729767268433
GITHUB: https://github.com/christianlempa
PATREON: https://www.patreon.com/christianlempa
MY EQUIPMENT: https://kit.co/christianlempa

Timestamps:

00:00 – Introduction
01:02 – Planning
02:24 – Server Rack types
05:12 – Additional Components
08:23 – Advertisement-*
09:16 – Devices in my Rack

________________
All links with “*” are affiliate links.

source

Continue Reading

Trending

Copyright © 2017 Zox News Theme. Theme by MVP Themes, powered by WordPress.