The increasing use of transformer-based large language models brings forward the challenge of processing long sequences. In document visual question answering (DocVQA), leading methods focus on the single-page setting, while documents can span hundreds of pages. We present GRAM, a method that seamlessly extends pre-trained single-page models to the multi-page setting, without requiring computationally-heavy pretraining. To do so, we leverage a single-page encoder for local page-level understanding, and enhance it with document-level designated layers and learnable tokens, facilitating the flow of information across pages for global reasoning. To enforce our model to utilize the newly introduced document-level tokens, we propose a tailored bias adaptation method. For additional computational savings during decoding, we introduce an optional compression stage using our C-Former model, which reduces the encoded sequence length, thereby allowing a tradeoff between quality and latency. Extensive experiments showcase GRAM's state-of-the-art performance on the benchmarks for multi-page DocVQA, demonstrating the effectiveness of our approach.

Test suites assess natural language processing models' performance on specific functionalities: cases of interest involving model robustness, fairness, or particular linguistic capabilities. They enable fine-grained evaluations of model aspects that would otherwise go unnoticed in standard evaluation datasets, but they do not address the problem of how to fix the failure cases. Previous work has explored functionality learning by fine-tuning models on suite data. While this improves performance on seen functionalities, it often does not generalize to unseen ones and can harm general performance. This paper analyses a fine-tuning-free approach to functionality learning. For each functionality in a suite, we generate a specification instruction that encodes it. We combine the obtained specification instructions to create specification-augmented prompts, which we feed to language models pre-trained on natural instruction data to generate suite predictions. A core aspect of our analysis is to measure the effect that including a set of specifications has on a held-out set of unseen, qualitatively different specifications. Our experiments across four tasks and models ranging from 80M to 175B parameters show that smaller models struggle to follow specification instructions. However, larger models (> 3B params.) can benefit from specifications and even generalize desirable behaviors across functionalities.

Corp. (PINKSHEETS: GCKO) -- announced today that one of their care giver participants has "looked in" on her mother while shopping. Her testimonial characterizes some of the benefits she is realizing from the CareBot and its remote monitoring and telepresence capabilities. GeckoSystems is a dynamic leader in the emerging Mobile Service Robot (MSR) industry revolutionizing their development and usage with their "Mobile Robot Solutions for Safety, Security, and Service ". "In the past I have not been able to'look in' on my elderly mother at home alone while I was out shopping. There was always a concern if she was okay during my absence, especially if she had some new health issue. Today, for the first time, I was able to watch her while grocery shopping at Ingles with the help of her CareBot. "I took my Dell netbook with me.

Corp. (PINKSHEETS: GCKO) -- -- announced today that they have further cost reduced their robot controller board, the GeckoSPIO, while improving ease of manufacturability and maintaining robust functionality. GeckoSystems is a dynamic leader in the emerging mobile robotics industry revolutionizing their development and usage with "Mobile Robot Solutions for Safety, Security, and Service ". "The GeckoSPIO is the critical interface between the robot's physical platform and higher AI functions. This interface provides a level of abstraction for the commands sent to, and the data sent from, the robot platform. The abstraction and hierarchal architecture the GeckoSPIO provides simplifies interacting with the platform and the real world for the high-level software, along with enabling a wide array of sensor fusion techniques. We are pleased that one of our recently hired electrical engineers has made these improvements," stated Mark Peele, Vice President, R&D, GeckoSystems.

CEO, Martin Spencer, addresses their 2018 progress in multiple markets. For over twenty years, GeckoSystems has dedicated itself to development of "AI Mobile Robot Solutions for Safety, Security and Service(tm)." We have further cost reduced the direct manufacturing costs of our proprietary, breakthrough SafePath! It is our AI enabled human quick sense and avoid of moving and/or unmapped obstacles that opens up the immense HME/DME rehab and power wheelchair markets to GeckoSystems for a first mover opportunity. We now believe that our collision free benefit can be added to most power wheelchairs for less than $2,000 MSRP, installed. "Recently, I had the opportunity to do some Delphi Market Research with one of Pride Mobility's largest US dealers and their master rehab power wheelchair technician.

Corp. (OTC-PINK: GOSY) (http://www.GeckoSystems.com/) announced today that their CEO, R. Martin Spencer, has filed RICO counterclaims against Neil T. Wallace seeking damages in excess of $ 47,000,000. For over twenty years, GeckoSystems has dedicated itself to development of "AI Mobile Robot Solutions for Safety, Security and Service … Social Media Analytics News

What does 2011 hold for the field of robotics? Plenty, if 2010 is any indication. This will not be the year that mobile, artificially intelligent robot nurses assume the responsibility of caring for the world's growing elderly population, but it does promise to be a pivotal time for the development of the underlying technology that will enable safe and reliable automated elder care, not to mention other services that robots are expected to perform in the coming decade. Thanks to a standardized platform introduced in 2010, roboticists can now collaborate as never before. Last May, Willow Garage, a Menlo Park, Calif., maker of robot hardware and software, released a test version of its personal robot platform.