In-context learning (ICL) is a powerful paradigm where large language models (LLMs) benefit from task demonstrations added to the prompt. Yet, selecting optimal demonstrations is not trivial, especially for complex or multi-modal tasks where input and output distributions differ. We hypothesize that forming task-specific representations of the input is key. In this paper, we propose a method to align representations of natural language questions and those of SQL queries in a shared embedding space. Our technique, dubbed MARLO - Metadata-Agnostic Representation Learning for Text-tO-SQL - uses query structure to model querying intent without over-indexing on underlying database metadata (i.e. tables, columns, or domain-specific entities of a database referenced in the question or query). This allows MARLO to select examples that are structurally and semantically relevant for the task rather than examples that are spuriously related to a certain domain or question phrasing. When used to retrieve examples based on question similarity, MARLO shows superior performance compared to generic embedding models (on average +2.9\%pt. in execution accuracy) on the Spider benchmark. It also outperforms the next best method that masks metadata information by +0.8\%pt. in execution accuracy on average, while imposing a significantly lower inference latency.

In "Star Wars," R2-D2 and C-3PO were involved in nothing less than vanquishing an evil empire. In reality, robots are bellhops. Aloft Hotels, which has a location in downtown Detroit, introduced its newest employee Tuesday -- Botlr. The three-foot robot, which looks like R2-D2 sans arms and legs, brings amenities to guests in the hotel lobby or right to their rooms. Moving on wheels, the machine has an enclosed bin that can hold anything from toothpaste to bottled water to whatever else the guest might desire.

Every chance we get, we post videos highlighting the adventures of MARLO, the University of Michigan's blissfully unaware bipedal robot. MARLO is totally "blind," without cameras, lidar, or anything else to show it where it's going. But the robot is still able to walk dynamically over a range of terrain that I think would be appropriate to call staggering. Varied terrain does indeed stagger MARLO on a regular basis, and it's probably fallen over more times on video than any robot we've ever seen. Professor Jessy Grizzle and his students have been challenging MARLO with increasingly difficult terrain, most recently at a location on the beautiful Ann Arbor campus called the "Wave Field," an "earth sculpture" created by artist Maya Lin.

Video Friday is your weekly selection of awesome robotics videos, collected by your laser-focused Automaton bloggers. We'll also be posting a weekly calendar of upcoming robotics events for the next two months; here's what we have so far (send us your events!): Let us know if you have suggestions for next week, and enjoy today's videos. Near Earth Autonomy, a Pittsburgh-based robotics and aerospace company (and CMU spin-off), has discovered the secret to flying a drone autonomously around obstacles without GPS: use a massive hexcopter that can haul around a Velodyne lidar. R1 is a new "personal humanoid" platform from the Italian Institute of Technology (IIT).

Like humans, robots often need to learn things the hard way. For super complex automatons, like Boston Dynamics' Atlas, that can mean understanding the complexities of human bullying. But in the case of this bumbling bot named MARLO, it's just something as simple as trying--and failing--to walk across a hilly field. MARLO is being designed by an engineering team at the University of Michigan. Although its walking capabilities look like MARLO is on the rough end of a day-long bender, MARLO's ability to traverse such uneven terrain would be a big step forward for robotics.

Video Friday is your weekly selection of awesome robotics videos, collected by your nebulous Automaton bloggers. We'll also be posting a weekly calendar of upcoming robotics events for the next two months; here's what we have so far (send us your events!): Let us know if you have suggestions for next week, and enjoy today's videos. Tend.ai makes "cloud robotics for machine tending" a reality. Our system uses artificial intelligence powered by cloud computing to read machines' displays and press machines' buttons just like a human would.

MARLO, the 3D bipedal robot that belongs to electrical engineering professor Jessy Grizzle and his team of students, is starting to really figure out this walking thing. Here, robotics PhD student Ross Hartley watches as MARLO demonstrate's her ability to conquer tough terrain. Image credit: Evan Dougherty, Michigan EngineeringANN ARBOR--An unsupported bipedal robot at the University of Michigan can now walk down steep slopes, through a thin layer of snow, and over uneven and unstable ground. The robot's feedback control algorithms should be able to help other two-legged robots as well as powered prosthetic legs gain similar capabilities. "The robot has no feeling in her tiny feet, but she senses the angles of her joints--for instance, her knee angles, hip angles and the rotation angle of her torso," said Jessy Grizzle, professor of electrical engineering and computer science and of mechanical engineering.

A bipedal robot has proven it doesn't need humans to help it maneuver uneven terrain or obstacles. Named MARLO, the machine has shown it can walk down steep slopes, through a thin layer of snow, and over uneven and unstable ground. The robot's feedback algorithms could be used to help program other two-legged robots as well as help prosthetic legs gain similar abilities. Dubbed MARLO, the cyborg is controlled with algorithms that allows the main controller to handle the forward and backward motion and the second to focus on side-to-side balance. This is the team's first robot that walk in all directions, which they believe can be used to help teach prosthetics to learn similar abilities Marlo is controlled with algorithms that enables the main controller to handle the forward and backward motion and the second to focus on side-to-side balance.

At first I was worried that this was going to be just another drone footage video with pompous music, but all of a sudden...that GIANT ROBOT EYEBALL BLIMP we saw in Germany last month shows up.

Aloft Hotel spokeswoman Ashley Chapman said the robot's performance of mundane tasks frees staff more time to interact with people. Detroit -- In "Star Wars," R2-D2 and C-3PO were involved in nothing less than vanquishing an evil empire. Aloft Hotels, which has a location in downtown Detroit, introduced its newest employee Tuesday -- Botlr. The three-foot robot, which looks like R2-D2 sans arms and legs, brings amenities to guests in the hotel lobby or right to their rooms. Moving on wheels, the machine has an enclosed bin that can hold anything from toothpaste to bottled water to whatever else the guest might desire.