Improving time-to-first-token (TTFT) is an essentially important objective in modern large language model (LLM) inference engines. Because optimizing TTFT directly results in higher maximal QPS and meets the requirements of many critical applications. However, boosting TTFT is notoriously challenging since it is purely compute-bounded and the performance bottleneck shifts from the self-attention to the MLP part. We present SpecPrefill, a training free framework that accelerates the inference TTFT for both long and medium context queries based on the following insight: LLMs are generalized enough to still preserve the quality given only a carefully chosen subset of prompt tokens. At its core, SpecPrefill leverages a lightweight model to speculate locally important tokens based on the context. These tokens, along with the necessary positional information, are then sent to the main model for processing. We evaluate SpecPrefill with a diverse set of tasks, followed by a comprehensive benchmarking of performance improvement both in a real end-to-end setting and ablation studies. SpecPrefill manages to serve Llama-3.1-405B-Instruct-FP8 with up to $7\times$ maximal end-to-end QPS on real downstream tasks and $7.66\times$ TTFT improvement during benchmarking.

Efficiently solving sparse linear systems $Ax=b$, where $A$ is a large, sparse, symmetric positive semi-definite matrix, is a core challenge in scientific computing, machine learning, and optimization. A major bottleneck in Gaussian elimination for these systems is fill-in, the creation of non-zero entries that increase memory and computational cost. Minimizing fill-in is NP-hard, and existing heuristics like Minimum Degree and Nested Dissection offer limited adaptability across diverse problem instances. We introduce \textit{ReFill}, a reinforcement learning framework enhanced by Graph Neural Networks (GNNs) to learn adaptive ordering strategies for fill-in minimization. ReFill trains a GNN-based heuristic to predict efficient elimination orders, outperforming traditional heuristics by dynamically adapting to the structure of input matrices. Experiments demonstrate that ReFill outperforms strong heuristics in reducing fill-in, highlighting the untapped potential of learning-based methods for this well-studied classical problem.

If you've been craving some help with cleaning, this Memorial Day Roomba sale may be up your alley. The machine not only cleans floors and carpets; its redesigned dock can automatically empty debris and refill it with mopping liquid. Use the exclusive code ENG425 to get iRobot's top-of-the-line combo cleaner for 974. The Roomba Combo J9 has an upgraded motor and a four-stage cleaning system that takes multiple passes across your floors and carpets. This latest model, which arrived only last fall, has dual rubber brushes for optimal suction and pressurized scrubbing.

Right now a code at Wellbots will get you iRobot's flagship Roomba Combo J9 for 400 off. The vacuum-mop hybrid robot, which only arrived last fall, automatically empties itself and refills the device's mopping liquid. That convenience doesn't come cheap, though. The unit usually sells for 1,399 but enter the coupon code ENGA400 at Wellbots and get it for 999 instead. Engadget's Devindra Hardawar gave the Combo J9 an 85 review score when it first came out, calling it a godsend for tired parents.

If you're looking to automate more of your home cleaning setup, iRobot's flagship Roomba Combo J9 is on sale for 400 off. The vacuum-mop hybrid robot, which only arrived last fall, has a redesigned dock that automatically empties debris and refills the device's mopping liquid. Take 400 off the vacuum-mop hybrid. The Combo J9, Engadget's pick for the best vacuuming-mopping robot, has an upgraded motor and four-stage cleaning system that takes multiple pass-throughs across your carpets and floors. This version also adds dual rubber brushes for better suction and pressurized scrubbing. Its motor automatically lifts the mop pads when it reaches carpets and rugs to help keep them dry.

I miss having clean floors. I've been using a variety of Roombas over the years to help out with vacuuming, but ever since my wife and I had our second child in 2022, mopping has become an afterthought. And really, vacuuming can only clean your floor so much. I missed the shimmer of a mopped hardwood floor and the smell of Murphy Oil cleaner lingering in the air. Instead, I've been living with even more toys, crumbs and an assortment of bodily waste (which three cats contribute to) on my flooring and carpets.

As more doctors and nurses leave the profession, providers are turning to AI technology to help bridge the coverage gap, especially among older Americans. As many as 55% of older adults fail to take their prescribed medications, according to data from the Department of Health and Human Services. Now a new artificial intelligence app aims to change that. "Together," a free iPhone app built on generative AI, is designed to help aging adults and their caregivers manage medications and other health care tasks. Using a smartphone camera, the person simply snaps a picture of a prescription bottle.

During physical human robot collaboration, it is important to be able to implement a time-varying interactive behaviour while ensuring robust stability. Admittance control and passivity theory can be exploited for achieving these objectives. Nevertheless, when the admittance dynamics is time-varying, it can happen that, for ensuring a passive and stable behaviour, some spurious dissipative effects have to be introduced in the admittance dynamics. These effects are perceived by the user and degrade the collaborative performance. In this paper we exploit the task redundancy of the manipulator in order to harvest energy in the null space and to avoid spurious dynamics on the admittance. The proposed architecture is validated by simulations and by experiments onto a collaborative robot.

In addition to launching the 12T smartphone series today, Xiaomi also announced that its Robot Vacuum X10 -- which debuted in China back in April -- is now headed to Europe. This is the company's most versatile robot vacuum to date: Not only can it vacuum with 4,000Pa suction and mop with pressurized dual rotating pads, but with its base station, it can automatically empty its trash, refill its water tank, wash its mops and dry them with heat as well. All told, the robot vacuum itself should require little maintenance; you'll just need to refill the base station's water reservoir and empty its waste water tank and dust bin from time to time. This isn't the first time Xiaomi released a robot vacuum with both mop-cleaning and mop-drying features; it did so for the Chinese market almost a year ago. With other models, you'd have to detach and wash the mop yourself which, let's be honest, isn't the most pleasant job in the world.

My dog died a few years ago. She was only about 7 years old and had a lot of health problems. But one thing I don't miss: my pharmacy bill. That's because Lavender (my wife chose the name, not me) was on no less than seven medications. And a day rarely went by without me a text message from my pharmacist at the CVS near me.