Linux color management has the same goal as the color management systems (CMS) for other operating systems, which is to achieve the best possible color reproduction throughout an imaging workflow from its source (camera, video, scanner, etc.), through imaging software (Digikam, darktable, RawTherapee, GIMP, Krita, Scribus, etc.), and finally onto an output medium (monitor, video projector, printer, etc.). In particular, color management attempts to enable color consistency across media and throughout a color-managed workflow. Linux color management relies on the use of accurate ICC (International Color Consortium) and DCP (DNG Color Profile) profiles describing the behavior of input and output devices, and color-managed applications that are aware of these profiles. These applications perform gamut conversions between device profiles and color spaces. Gamut conversions, based on accurate device profiles, are the essence of color management. Historically, color management was not an initial design consideration of the X Window System on which much of Linux graphics support rests, and thus color-managed workflows have been somewhat more challenging to implement on Linux than on other OS's such as Microsoft Windows or macOS. This situation is now being progressively remedied, and color management under Linux, while functional, has not yet acquired mature status. Although it is now possible to obtain a consistent color-managed workflow under Linux, certain problems still remain: The absence of a central user control panel for color settings. Some hardware devices for color calibration lack Linux drivers, firmware or accessory data. Since ICC color profiles are written to an open specification, they are compatible across operating systems. Hence, a profile produced on one OS should work on any other OS given the availability of the necessary software to read it and perform the gamut conversions. This can be used as a workaround for the lack of support for certain spectrophotometers or colorimeters under Linux: one can simply produce a profile on a different OS and then use it in a Linux workflow. Additionally, certain hardware, such as most printers and certain monitors, can be calibrated under another OS and then used in a fully color-managed workflow on Linux. The popular Ubuntu Linux distribution added initial color management in the 11.10 release (the "Oneiric Ocelot" release). == Requirements for a color-managed workflow == Accurate device profiles obtained with source or output characterization software. Correctly loaded video card lookup tables (LUTs) (or monitor profiles that do not require LUT adjustments). Color-managed applications that are configured to use a correct monitor profile and input/output profiles, with support for control over the rendering intent and black point compensation. Calibration and profiling requires: for input devices (scanner, camera, etc.) a color target which the profiling software will compare to the manufacturer-provided color values of the target. or for output devices (monitor, printer, etc.) a reading with a specific device (spectrophotometer, colorimeter or spectrocolorimeter) of the color patch values and comparing the measured values against the values originally sent for output. === Monitor calibration and profiling === One of the critical elements in any color-managed workflow is the monitor, because, at one step or another, handling and making color adaptation through imaging software is required for most images, thus the ability of the monitor to present accurate colors is crucial. Monitor color management consists of calibration and profiling. The first step, calibration, is done by adjusting the monitor controls and the output of the graphics card (via calibration curves) to match user-definable characteristics, such as brightness, white point and gamma. The calibration settings are stored in a .cal file. The second step, profiling (characterization), involves measuring the calibrated display's response and recording it in a color profile. The profile is stored in an .icc file ("ICC file"). For convenience, the calibration settings are usually stored together with the profile in the ICC file. Note that .icm files are identical to .icc files - the difference is only in the name. Seeing correct colors requires using a monitor profile-aware application, together with the same calibration used when profiling the monitor. Calibration alone does not yield accurate colors. If a monitor was calibrated before it was profiled, the profile will only yield correct colors when used on the monitor with the same calibration (the same monitor control adjustments and the same calibration curves loaded into the video card's lookup table). macOS has built-in support for loading calibration curves and installing a system-wide color profile. Windows 7 onward allows loading calibration curves, though this functionality must be enabled manually. Linux and older versions of Windows require using a standalone LUT loader. === Device profiles === ICC profiles are cross-platform and can thus be created on other operating systems and used under Linux. Monitor profiles, however, require some additional attention. Since a monitor profile depends both on the monitor itself and on the video card, a monitor profile should only be used with the same monitor and video card with which it was created. The monitor settings should not be adjusted after creating the profile. In addition, since most calibration software use LUT adjustments during calibration, the corresponding LUTs must be loaded every time the display server (X11, Wayland) is started (e.g. with every graphical login). In the unlikely case of a colorimeter being unsupported by Linux, a profile created under Windows or macOS can be used under Linux. === Display-channel lookup tables === There are two approaches to loading display channel LUTs: Create a profile that does not modify video card LUTs and thus does not require LUTs be loaded later on. Ideally, this approach would rely on DDC-capable monitors—the internal monitor settings of which are set via calibration software. Unfortunately, monitors capable of making these adjustments through DDC are not common and are generally expensive. There is only one calibration software on Linux that can interact with a DDC monitor. For mainstream monitors, a couple of options exist: BasICColor software, which works with most colorimeters on the market, allows one to adjust display output via the monitor interface, and then to choose a "Profile, do not calibrate" option. By doing this, one can create a profile that does not require video card LUT adjustments. For EyeOne devices, EyeOne Match allows the user to calibrate to "Native" gamma and white point targets, which results in the LUT adjustment curves displayed after the calibration as a simple, linear 1:1 mapping (a straight line from corner to corner). Both BasICColor and EyeOne Match do not presently run under Linux but they are capable of creating a profile that does not require LUT adjustments. Use an LUT loader to actually load the LUT adjustments contained within the profile prepared during calibration. According to the documentation, these loaders do not modify the video card LUT by itself, but achieve the same type of adjustment by modifying the X server gamma ramp. Loaders are available for Linux distributions that use X.org or XFree86—the two most popular X servers on Linux. Other X servers are not guaranteed to work with the currently available loaders. There are two LUT loaders available for Linux: Xcalib is one such loader, and although it is a command-line utility, it is quite easy to use. dispwin is a part of Argyll CMS. If, for any reason, the LUT cannot be loaded, it is still recommended to go through the initial stages of calibration where a user is asked by calibration software to make some manual adjustments to the monitor, as this will often improve display linearity and also provide information on its color temperature. This is especially recommended for CRT monitors. === Color-managed applications === In ICC-aware applications, it is important to make sure the correct profiles are assigned to devices, mainly to the monitor and the printer. Some Linux applications can auto-detect the monitor profile, while others requires that it is specified manually. Although there is no designated place to store device profiles on Linux, /usr/share/color/icc/ has become the de facto standard. Most applications running under WINE have not been fully tested for color accuracy. While 8-bpp programs can have some color resolution difficulties due to depth conversion errors, colors in higher-depth applications should be accurate, as long as those programs perform their gamut conversions based on the same monitor profile as that used for loading the LUT, granted that the corresponding LUT adjustments are loaded. == List of color-managed applications == darktabl
Colloquis
Colloquis, previously known as ActiveBuddy and Conversagent, was a company that created conversation-based interactive agents originally distributed via instant messaging platforms. The company had offices in New York, New York, and Sunnyvale, California. == History == Founded in 2000, the company was the brainchild of Robert Hoffer, Timothy Kay, and Peter Levitan. The idea for interactive agents (also known as Internet bots) came from the team's vision to add functionality to increasingly popular instant messaging services. The original implementation took shape as a word-based adventure game but quickly grew to include a wide range of database applications, including access to news, weather, stock information, movie times, Yellow Pages listings, and detailed sports data, as well as a variety of tools (calculators, translator, etc.). These various applications were bundled into one entity and launched as SmarterChild in 2001. SmarterChild acted as a showcase for the quick data access and possibilities for fun conversation that the company planned to turn into customized, niche-specific products. The rapid success of SmarterChild led to targeted promotional products for Radiohead, Austin Powers, The Sporting News, and others. ActiveBuddy sought to strengthen its hold on the interactive agent market for the future by filing for, and receiving, a controversial patent on their creation in 2002. The company also released the BuddyScript SDK, a free developer kit that allow programmers to design and launch their own interactive agents using ActiveBuddy's proprietary scripting language, in 2002. Ultimately, however, the decline in ad spending in 2001 and 2002 led to a shift in corporate strategy towards business focused Automated Service Agents, building products for clients including Cingular, Comcast and Cox Communications. The company subsequently changed its name from ActiveBuddy to Conversagent in 2003, and then again to Colloquis in 2006. Colloquis was purchased by Microsoft in October 2006.
Verbot
The Verbot (short for Verbal-Robot) was a chatbot program and artificial intelligence software development kit (SDK) designed for Windows and web platforms. == Early beginning == The origin of verbot traces back to Michael Mauldin's research during his time as a graduate student and post-doctoral fellow at Carnegie Mellon University. The creative foundation also stems from Peter Plantec's work in personality psychology and art direction. === Historic outline === In 1994, Michael Loren Mauldin, founder of Lycos, Inc., developed a prototype chatbot, Julia, which competed in the internationally known Turing test, for the coveted Loebner Prize. The Turing test matches computer scientist judges against machines to see if they can distinguish a computer from a real human. Julia was refined and developed, and in 1997, Dr. Mauldin and Peter Plantec, a clinical psychologist and animator, formed Virtual Personalities, Inc. (now Conversive, Inc.) in order to create a virtual human interface that would incorporate real-time animation as well as speech and natural language processing. The initial release, a stand-alone virtual person called Sylvie, was beta-tested to the public. This release was well received, and finally, after several versions, the production release (deemed version 3) of the Verbally Enhanced Software Robot, or Verbot, was deployed in fall 2000. The grandfather of all Verbots is Rog-O-Matic, which, although it could not talk, could and did explore a virtual world. Julia has been active on the internet in one form or another since 1989. A close cousin of Julia is Lycos, a robot that explores the World Wide Web and answers questions about it. Sylvie was the first Verbot with a face and a voice. Sylvie was the first Virtual Human with advanced, flexible interfacing capability. === Beginnings === The Virtual Personalities story goes back to 1978, where Mauldin was attending Rice University. Fascinated by the idea of ELIZA, he proceeded to write a program called "PET" for his 8 kilobyte Commodore PET Computer. PET included simple induction as a way to post new information, for example: Subject: I like my friend (later) Subject: I like food. PET: I have heard that food is your friend. Meanwhile, Plantec was separately designing a personality for "Entity", a theoretical virtual human that would interact comfortably with humans without pretending to be one. At that time the technology was not advanced enough to realize Entity. Mauldin got so involved with this that he majored in Computer Science and minored in Linguistics. === Rogue === In the late seventies and early eighties, a popular computer game at universities was Rogue, an implementation of Dungeons and Dragons where the player would descend 26 levels in a randomly created dungeon, fighting monsters, gathering treasure, and searching for the elusive "Amulet of Yendor". Mauldin was one of four grad students who devoted a large amount of time to building a program called "Rog-O-Matic" capable of retrieving the amulet and emerging victorious from the dungeon. === TinyMUD === In 1989, when James Aspnes at Carnegie Mellon created the first TinyMUD (a descendant of MUD and AberMUD), Mauldin was one of the first to create a computer player that would explore the text-based world of TinyMUD. But his first robot, Gloria, gradually accreted more and more linguistic ability, to the point that it could pass the "unsuspecting" Turing test. In this version of the test, the human has no reason to suspect that one of the other occupants of the room is controlled by a computer, and so is more polite and asks fewer probing questions. The second generation of Mauldin's TinyMUD robots was Julia, created on Jan. 8, 1990. Julia slowly developed into a more and more capable conversational agent, and assumed useful duties in the TinyMUD world, including tour guide, information assistant, note-taker, and message-relayer. She could even play the card game hearts along with the other human players. In 1991, Julia attended the first Loebner Prize contest in Boston, Massachusetts. Although she only finished third, she was ranked by one judge as more human than one of the human confederates, winning a coveted certificate of humanness in the world's first restricted Turing test. Julia continued to log in to various TinyMUD's and TinyMucks for the next seven years, and chatted with hundreds of people a month over the internet. === Lycos === Julia's job was to explore a virtual world consisting of pages of textual descriptions, with links between them, and to construct an internal map of that world and answer questions about it (including path information such as the shortest route from one room to another, and matching information, such as which rooms contained a certain kind of object or textual description). It was therefore only a very short cognitive leap from Julia to Lycos, another robotic agent that explores a virtual world made of hyperlinked pages of text, and which answers questions about those pages. Sylvie was born and her abilities were expanded greatly to include interfacing with computers and control systems via her serial ports. === Sylvie === Sylvie was the first intelligent animated virtual human. She was designed both as a conversation agent and as a virtual human interface that would form a bridge between the two. She became more popular as a conversation agent, but her designers believe she serves as a prototype for future virtual human interface design that will help us all cope with the increasing complexity of technology. As an aside, Plantec noticed that a large number of Sylvies have been sold in Southeast Asia. Upon investigation, he found out that students had discovered a "test" mode that would allow them to type in English sentences that Sylvie would pronounce in her somewhat stylized English. == Ownership == In 1997, Dr. Mauldin and Peter Plantec formed Virtual Personalities, Inc. to create Natural Language Processing solutions for companies. In 2001 Virtual Personalities, Inc. became Conversive, Inc. to reflect the focus on providing Customer Service and Marketing to the Enterprise Market. In late 2012 Avaya, Inc. acquired Conversive's assets including Verbots. == Verbot versions == The Verbot 4 version was created and released in 2004. In 2005 Version 4.1 of the Verbot Software was released with many feature enhancements and bug fixes, including built-in support for embedding C# code in outputs and conditionals. In early 2006 Conversive launched Verbots Online allowing Verbot 4 users to upload their knowledge and show off their bots to the world. In 2009 Version 5 was released, completely free and fully featured. In early 2012 the last version of Verbot, 5.0.1.2, was released to the general public with support for Windows 7. Later in 2012 Verbots Online completely shut down. == Verbots today == Verbots.com, its community of users, and its forums no longer exist, but the software and users can still be found. There has been no active development since the early 2012 release of Verbot 5.0.1.2.
International Conference on Language Resources and Evaluation
The International Conference on Language Resources and Evaluation is an international conference organised by the ELRA Language Resources Association every other year (on even years) with the support of institutions and organisations involved in Natural language processing. The series of LREC conferences was launched in Granada in 1998. == History of conferences == The survey of the LREC conferences over the period 1998-2013 was presented during the 2014 conference in Reykjavik as a closing session. It appears that the number of papers and signatures is increasing over time. The average number of authors per paper is higher as well. The percentage of new authors is between 68% and 78%. The distribution between male (65%) and female (35%) authors is stable over time. The most frequent technical term is "annotation", then comes "part-of-speech". == The LRE Map == The LRE Map was introduced at LREC 2010 and is now a regular feature of the LREC submission process for both the conference papers and the workshop papers. At the submission stage, the authors are asked to provide some basic information about all the resources (in a broad sense, i.e. including tools, standards and evaluation packages), either used or created, described in their papers. All these descriptors are then gathered in a global matrix called the LRE Map. This feature has been extended to several other conferences.
Ayoba
Ayoba is an African communication platform developed in South Africa. It is owned by Progressive Tech Holdings in Mauritius and managed by SIMFY Africa. Launched on May 4, 2019, as of April 2024, it has over 35 million active users. == History == Ayoba was first published on Google Play in February 2019. Its first marketing campaign and brand launch took place in Cameroon on May 4, 2019. In June 2019, the platform introduced its first eight channels. In November 2019, the platform reached one million active users, which increased to two million by June 2020. Subsequently, ayoba expanded its services, including the launch of games for Android in February 2020, Momo (Mobile Money) in Cameroon in May 2020, and MicroApps in May 2020. It also launched music and voice and video calling features in 12 territories in August 2020. The first version of ayoba for iOS was released in September 2020. In December of the same year, games and Messaging 2.0 were launched on the platform. In November 2020, it won Best Mobile Application at the African Digital Awards. In 2021, it won OTT Brand of the Year at the Marketing World Awards in Ghana. In December 2022, it received Top Innovative Technology and Telecom Product of the Year at the National Communications Awards in December 2022. In June 2023 ayoba partnered with BoomPlay and as of April 2024, it had 35 million monthly active users. Ayoba has partnered with Jumia Ghana to offer exclusive deals to users. Ayoba users can get a 10% discount on selected Jumia purchases through the app, with no data charges for MTN users. This partnership aims to make online shopping more affordable and accessible by integrating Jumia's offers into the ayoba app. Ayoba supports over 35 million users across Africa and provides services in 22 languages. To access the deals, users can download the ayoba app from the Google Play Store, iOS Store, or the official website. == Platform features == Chat, Call and Share: ayoba enables instant messaging, voice notes, picture sharing, and file sharing with contacts, even if they do not have the app installed. The app supports voice and video calls on both Android and iOS, as well as group chats, help channel and SMS continuity (non ayoba users receive messages as SMS, their responses appear in the ayoba app). Music: ayoba offers a free music player with daily updates on international and African music. Users can find playlists for different genres. Games: ayoba provides a selection of interactive games, including action, adventure, and children's games available on both Android and iOS. Mobile Money Transfers: In certain territories, ayoba supports mobile money transfers using MTN Mobile Money (MoMo) for transactions within the app. MicroApps: ayoba features individual MicroApps within the platform that offer content and services, including streaming channels, podcasts, and specialized apps. The availability of these apps may vary by country. == Operations == ayoba primarily focuses on the following territories: Nigeria, Cameroon, South Africa, Ghana, Côte d'Ivoire, Uganda, Republic of Congo, Benin, Zambia, Tanzania, Kenya, Senegal, Togo, Guinea Bissau, Guinea Conakry, Sudan, South Sudan, and Liberia. The company operates from its offices in Cape Town and Johannesburg, South Africa. David Gillaranz served as the CEO from 2019 to 2021, and Burak Akinci has been the CEO since 2021.
Enterprise resource planning
Enterprise resource planning (ERP) is the integrated management of main business processes, often in real time and mediated by software and technology. ERP is usually referred to as a category of business management software—typically a suite of integrated applications—that an organization can use to collect, store, manage and interpret data from many business activities. The finance module in particular is essential to a suite of applications meeting the definition of an ERP system. The finance module provides the system of record for the organisation; recording the commercial impact of the business operations in the General Ledger. ERP systems can be local-based or cloud-based. Cloud-based applications have grown rapidly since the early 2010s due to the increased efficiencies arising from information being readily available from any location with Internet access. However, ERP differs from integrated business management systems by including planning all resources that are required in the future to meet business objectives. This includes plans for getting suitable staff and manufacturing capabilities for future needs. ERP provides an integrated and continuously updated view of core business processes, typically using a shared database managed by a database management system. ERP systems track business resources—cash, raw materials, production capacity—and the status of business commitments: orders, purchase orders, and payroll. The applications that make up the system share data across various departments (manufacturing, purchasing, sales, accounting, etc.) that provide the data. ERP facilitates information flow between all business functions and manages connections to outside stakeholders. Estimates of the size of the global ERP market range between USD $78 and $81 billion in 2026 . Though early ERP systems focused on large enterprises, smaller enterprises increasingly use ERP systems. The ERP system integrates varied organizational systems and facilitates error-free transactions and production, thereby enhancing the organization's efficiency. However, developing an ERP system differs from traditional system development. ERP systems run on a variety of computer hardware and network configurations, typically using a database as an information repository. == Origin == Business and technology research and advisory firm Gartner is credited for first using the acronym ERP in the 1990s. The term captured a functional extension of two manufacturing-based concepts, material requirements planning (MRP) and manufacturing resource planning (MRP II). Without replacing these terms, ERP came to represent a larger whole that reflected the evolution of application integration beyond manufacturing. Not all ERP packages are developed from a manufacturing core; ERP vendors variously began assembling their packages with finance-and-accounting, maintenance, and human-resource components. By the mid-1990s ERP systems addressed all core enterprise functions. Governments and non–profit organizations also began to use ERP systems. An "ERP system selection methodology" is a formal process for selecting an enterprise resource planning (ERP) system. Existing methodologies include: Kuiper's funnel method, Dobrin's three-dimensional (3D) web-based decision support tool, and the Clarkston Potomac methodology. == Expansion == ERP systems experienced rapid growth in the 1990s. Because of the year 2000 problem many companies took the opportunity to replace their old systems with ERP. ERP systems initially focused on automating back office functions that did not directly affect customers and the public. Front office functions, such as customer relationship management (CRM), dealt directly with customers, or e-business systems such as e-commerce and e-government—or supplier relationship management (SRM) became integrated later, when the internet simplified communicating with external parties. "ERP II" was coined in 2000 in an article by Gartner Publications entitled ERP Is Dead—Long Live ERP II. It describes web–based software that provides real–time access to ERP systems to employees and partners (such as suppliers and customers). The ERP II role expands traditional ERP resource optimization and transaction processing. Rather than just manage buying, selling, etc.—ERP II leverages information in the resources under its management to help the enterprise collaborate with other enterprises. ERP II is more flexible than the first generation ERP. Rather than confine ERP system capabilities within the organization, it goes beyond the corporate walls to interact with other systems. Enterprise application suite is an alternate name for such systems. ERP II systems are typically used to enable collaborative initiatives such as supply chain management (SCM), customer relationship management (CRM) and business intelligence (BI) among business partner organizations through the use of various electronic business technologies. The large proportion of companies are pursuing a strong managerial targets in ERP system instead of acquire an ERP company. Developers now make more effort to integrate mobile devices with the ERP system. ERP vendors are extending ERP to these devices, along with other business applications, so that businesses don't have to rely on third-party applications. As an example, the e-commerce platform Shopify was able to make ERP tools from Microsoft and Oracle available on its app in October 2021. Technical stakes of modern ERP concern integration—hardware, applications, networking, supply chains. ERP now covers more functions and roles—including decision making, stakeholders' relationships, standardization, transparency, globalization, etc. == Functional areas == An ERP system covers the following common functional areas. In many ERP systems, these are called and grouped together as ERP modules: Financial accounting: general ledger, fixed assets, payables including vouchering, matching and payment, receivables and collections, cash management, financial consolidation Management accounting: budgeting, costing, cost management, activity based costing, billing, invoicing (optional) Human resources: recruiting, training, rostering, payroll, benefits, retirement and pension plans, diversity management, retirement, separation Manufacturing: engineering, bill of materials, work orders, scheduling, capacity, workflow management, quality control, manufacturing process, manufacturing projects, manufacturing flow, product life cycle management Order processing: order to cash, order entry, credit checking, pricing, available to promise, inventory, shipping, sales analysis and reporting, sales commissioning Supply chain management: supply chain planning, supplier scheduling, product configurator, order to cash, purchasing, inventory, claim processing, warehousing (receiving, putaway, picking and packing) Project management: project planning, resource planning, project costing, work breakdown structure, billing, time and expense, performance units, activity management Customer relationship management (CRM): sales and marketing, commissions, service, customer contact, call center support – CRM systems are not always considered part of ERP systems but rather business support systems (BSS) Supplier relationship management (SRM): suppliers, orders, payments. Data services: various "self-service" interfaces for customers, suppliers or employees Management of school and educational institutes. Contract management: creating, monitoring, and managing contracts, reducing administrative burdens and minimising legal risks. These modules often feature contract templates, electronic signature capabilities, automated alerts for contract milestones, and advanced search functionality. === GRP – ERP use in government === Government resource planning (GRP) is the equivalent of an ERP for the public sector and an integrated office automation system for government bodies. The software structure, modularization, core algorithms and main interfaces do not differ from other ERPs, and ERP software suppliers manage to adapt their systems to government agencies. Both system implementations, in private and public organizations, are adopted to improve productivity and overall business performance in organizations, but comparisons (private vs. public) of implementations shows that the main factors influencing ERP implementation success in the public sector are cultural. == Best practices == Most ERP systems incorporate best practices. This means the software reflects the vendor's interpretation of the most effective way to perform each business process. Systems vary in how conveniently the customer can modify these practices. Use of best practices eases compliance with requirements such as International Financial Reporting Standards, Sarbanes–Oxley, or Basel II. They can also help comply with de facto industry standards, such as electronic funds transfer. This is because the procedure can be readily
Language model benchmark
A language model benchmark is a standardized test designed to evaluate the performance of language models on various natural language processing tasks. These tests are intended for comparing different models' capabilities in areas such as language understanding, generation, and reasoning. Benchmarks generally consist of a dataset and corresponding evaluation metrics. The dataset provides text samples and annotations, while the metrics measure a model's performance on tasks like answering questions, text classification, and machine translation. These benchmarks are developed and maintained by academic institutions, research organizations, and industry players to track progress in the field. In addition to accuracy, the metrics can include throughput, energy efficiency, bias, trust, and sustainability. == Overview == === Types === Benchmarks may be described by the following adjectives, not mutually exclusive: Classical: These tasks are studied in natural language processing, even before the advent of deep learning. Examples include the Penn Treebank for testing syntactic and semantic parsing, as well as bilingual translation benchmarked by BLEU scores. Question answering: These tasks have a text question and a text answer, often multiple-choice. They can be open-book or closed-book. Open-book QA resembles reading comprehension questions, with relevant passages included as annotation in the question, in which the answer appears. Closed-book QA includes no relevant passages. Closed-book QA is also called open-domain question-answering. Before the era of large language models, open-book QA was more common, and understood as testing information retrieval methods. Closed-book QA became common since GPT-2 as a method to measure knowledge stored within model parameters. Omnibus: An omnibus benchmark combines many benchmarks, often previously published. It is intended as an all-in-one benchmarking solution. Reasoning: These tasks are usually in the question-answering format, but are intended to be more difficult than standard question answering. Multimodal: These tasks require processing not only text, but also other modalities, such as images and sound. Examples include OCR and transcription. Agency: These tasks are for a language-model–based software agent that operates a computer for a user, such as editing images, browsing the web, etc. Adversarial: A benchmark is "adversarial" if the items in the benchmark are picked specifically so that certain models do badly on them. Adversarial benchmarks are often constructed after state of the art (SOTA) models have saturated (achieved 100% performance) a benchmark, to renew the benchmark. A benchmark is "adversarial" only at a certain moment in time, since what is adversarial may cease to be adversarial as newer SOTA models appear. Public/Private: A benchmark might be partly or entirely private, meaning that some or all of the questions are not publicly available. The idea is that if a question is publicly available, then it might be used for training, which would be "training on the test set" and invalidate the result of the benchmark. Usually, only the guardians of the benchmark have access to the private subsets, and to score a model on such a benchmark, one must send the model weights, or provide API access, to the guardians. The boundary between a benchmark and a dataset is not sharp. Generally, a dataset contains three "splits": training, test, and validation. Both the test and validation splits are essentially benchmarks. In general, a benchmark is distinguished from a test/validation dataset in that a benchmark is typically intended to be used to measure the performance of many different models that are not trained specifically for doing well on the benchmark, while a test/validation set is intended to be used to measure the performance of models trained specifically on the corresponding training set. In other words, a benchmark may be thought of as a test/validation set without a corresponding training set. Conversely, certain benchmarks may be used as a training set, such as the English Gigaword or the One Billion Word Benchmark, which in modern language is just the negative log-likelihood loss on a pretraining set with 1 billion words. Indeed, the distinction between benchmark and dataset in language models became sharper after the rise of the pretraining paradigm, whereby a model is first trained on massive, unlabeled datasets to learn general language patterns, syntax, and knowledge (pretraining), and the base model is then adapted to specific, downstream tasks using smaller, labeled datasets (fine-tuning). === Lifecycle === Generally, the life cycle of a benchmark consists of the following steps: Inception: A benchmark is published. It can be simply given as a demonstration of the power of a new model (implicitly) that others then picked up as a benchmark, or as a benchmark that others are encouraged to use (explicitly). Growth: More papers and models use the benchmark, and the performance on the benchmark grows. Maturity, degeneration or deprecation: A benchmark may be saturated, after which researchers move on to other benchmarks. Progress on the benchmark may also be neglected as the field moves to focus on other benchmarks. Renewal: A saturated benchmark can be upgraded to make it no longer saturated, allowing further progress. === Construction === Like datasets, benchmarks are typically constructed by several methods, individually or in combination: Web scraping: Ready-made question-answer pairs may be scraped online, such as from websites that teach mathematics and programming. Conversion: Items may be constructed programmatically from scraped web content, such as by blanking out named entities from sentences, and asking the model to fill in the blank. This was used for making the CNN/Daily Mail Reading Comprehension Task. Crowd sourcing: Items may be constructed by paying people to write them, such as on Amazon Mechanical Turk. This was used for making the MCTest. === Evaluation === Generally, benchmarks are fully automated. This limits the questions that can be asked. For example, with mathematical questions, "proving a claim" would be difficult to automatically check, while "calculate an answer with a unique integer answer" would be automatically checkable. With programming tasks, the answer can generally be checked by running unit tests, with an upper limit on runtime. The benchmark scores are of the following kinds: For multiple choice or cloze questions, common scores are accuracy (frequency of correct answer), precision, recall, F1 score, etc. pass@n: The model is given n {\displaystyle n} attempts to solve each problem. If any attempt is correct, the model earns a point. The pass@n score is the model's average score over all problems. k@n: The model makes n {\displaystyle n} attempts to solve each problem, but only k {\displaystyle k} attempts out of them are selected for submission. If any submission is correct, the model earns a point. The k@n score is the model's average score over all problems. cons@n: The model is given n {\displaystyle n} attempts to solve each problem. If the most common answer is correct, the model earns a point. The cons@n score is the model's average score over all problems. Here "cons" stands for "consensus" or "majority voting". The pass@n score can be estimated more accurately by making N > n {\displaystyle N>n} attempts, and use the unbiased estimator 1 − ( N − c n ) ( N n ) {\displaystyle 1-{\frac {\binom {N-c}{n}}{\binom {N}{n}}}} , where c {\displaystyle c} is the number of correct attempts. For less well-formed tasks, where the output can be any sentence, there are the following commonly used scores including BLEU ROUGE, METEOR, NIST, word error rate, LEPOR, CIDEr, and SPICE. === Issues === error: Some benchmark answers may be wrong. ambiguity: Some benchmark questions may be ambiguously worded. subjective: Some benchmark questions may not have an objective answer at all. This problem generally prevents creative writing benchmarks. Similarly, this prevents benchmarking writing proofs in natural language, though benchmarking proofs in a formal language is possible. open-ended: Some benchmark questions may not have a single answer of a fixed size. This problem generally prevents programming benchmarks from using more natural tasks such as "write a program for X", and instead uses tasks such as "write a function that implements specification X". inter-annotator agreement: Some benchmark questions may be not fully objective, such that even people would not agree with 100% on what the answer should be. This is common in natural language processing tasks, such as syntactic annotation. shortcut: Some benchmark questions may be easily solved by an "unintended" shortcut. For example, in the SNLI benchmark, having a negative word like "not" in the second sentence is a strong signal for the "Contradiction" category, regardless of what the se