A confusion network (sometimes called a word confusion network or informally known as a sausage) is a natural language processing method that combines outputs from multiple automatic speech recognition or machine translation systems. Confusion networks are simple linear directed acyclic graphs with the property that each a path from the start node to the end node goes through all the other nodes. The set of words represented by edges between two nodes is called a confusion set. In machine translation, the defining characteristic of confusion networks is that they allow multiple ambiguous inputs, deferring committal translation decisions until later stages of processing. This approach is used in the open source machine translation software Moses and the proprietary translation API in IBM Bluemix Watson.
BiP (software)
BiP is a freeware instant messaging application developed by Lifecell Ventures Cooperatief U.A., a subsidiary of Turkcell incorporated in the Netherlands. It allows users to send text messages, voice messages and video calling, and it can be downloaded from the App Store, Google Play, and Huawei AppGallery. BiP has over 53 million users worldwide, and was first released in 2013. == Functions == BiP is a secure, and free communication platform. BiP allows making video and audio calls, allows sharing images, videos and location. BiP includes instant translations to 106 languages and exchange rates. President Erdoğan's Communications Office opposed WhatsApp's enforcement of its updated privacy policy and announced that Erdoğan left WhatsApp and opened an account in Telegram and BiP. The Turkish Ministry of National Defense has announced that it will move information groups to BiP for the same reason. == Others == Banglalink announced a BiP messenger partnership in Bangladesh The Communications Office of President Erdoğan opposed WhatsApp's enforcement of its updated privacy policy and announced that Erdoğan left WhatsApp and opened an account in Telegram and BiP. The Turkish Ministry of National Defense has announced that it will move information groups to BiP for the same reason. The CEO of BiP is Burak Akinci. The number of downloads of the app is 80 million globally.
Social media marketing
Social media marketing is the use of social media platforms and websites to promote a product or service. Although the terms e-marketing and digital marketing are still dominant in academia, social media marketing is becoming more popular for practitioners and researchers. Social media platforms such as Facebook, LinkedIn, Instagram, and Twitter, among others, have built-in data analytics tools that companies can use to track the progress, success, and engagement of social media marketing campaigns. Companies address a range of stakeholders through social media marketing, including current and potential customers, current and potential employees, journalists, bloggers, and the general public. On a strategic level, social media marketing includes the management of a marketing campaign, governance, setting the scope (e.g. more active or passive use) and the establishment of a firm's desired social media "culture" and "tone". Firms that use social media marketing can allow customers and Internet users to post user-generated content (e.g., online comments, product reviews, etc.), also known as "earned media", rather than use marketer-prepared advertising copy. == Purposes and tactics == Social media may be employed in marketing as a communications tool that makes companies accessible to those who are interested in their product and visible to those who are not familiar with their products. It is used by companies to create buzz, learn from customers, and target them. Of the top 10 factors that correlate with a strong Google organic search, seven are social media-dependent. This means that if brands with little to no social media presence tend to show up less on Google searches. While platforms such as Twitter, Facebook and—in the past—Google+ have a larger number of monthly users, the visual media-sharing-based mobile platforms garner a higher interaction rate in comparison, and have registered the fastest growth, and have changed the ways in which consumers engage with brand content. Instagram has an interaction rate of 1.46% with an average of 130 million users monthly as opposed to Twitter, which has a .03% interaction rate with an average of 210 million monthly users. Unlike traditional media that are often cost-prohibitive to many companies, a social media strategy does not require significant financial investment. To this end, companies make use of platforms such as Facebook, Twitter, YouTube, TikTok and Instagram to reach audiences much wider than through traditional print, television, or radio advertisements alone at a fraction of the cost, as most social networking sites can be used at little or no cost (however, some websites charge companies for premium services). This has changed the ways that companies approach and interact with customers, as a substantial percentage of consumer interactions are now being carried out over online platforms with much higher visibility. Customers can post reviews of products and services, rate customer service, and ask questions or voice concerns directly to companies through social media platforms. According to Measuring Success, over 80% of consumers use the web to research products and services. Thus social media marketing is also used by businesses in order to build relationships of trust with consumers. To this aim, companies may hire personnel to specifically handle these social media interactions, who usually report under the title of online community managers. Handling these interactions in a satisfactory manner can result in an increase of consumer trust. To both this aim and to fix the public's perception of a company, three steps are taken in order to address consumer concerns: Identifying the extent of the social chatter Engaging the influencers to help Developing a proportional response == Strategies == === Passive approach === Social media can be a useful source of market information and a way to hear customers' perspectives. Blogs, content communities, and forums are platforms where individuals share their reviews and recommendations of brands, products, and services. Businesses are able to tap into and analyze customer voices and feedback generated in social media for marketing purposes. In this sense, social media is a relatively inexpensive source of market intelligence which can be used by marketers and managers to track and respond to consumer-identified problems and detect market opportunities. === Active approach === Social media can be used as a public relations tool, a direct marketing tool, and a communication channel to target very specific audiences, with social media influencers and social media personalities as effective customer engagement tools. This tactic is widely known as influencer marketing, which gives brands the opportunity to reach their target audience via a group of selected influencers advertising their product or service. Brands were projected to spend up to $15 billion on influencer marketing by 2022, per Business Insider Intelligence estimates, based on Mediakix data. The use of customer influencers, such as popular bloggers, can be an efficient and cost-effective method to launch new products or services. == Engagement == Engagement with the social web means that customers and stakeholders are active participants rather than passive spectators. An example of these are consumer advocacy groups and groups that criticize companies (e.g., lobby groups or advocacy organizations). The use of Social media in a business or political context allows people to express and share opinions about a company's products, services or business practices, or a government's actions. On social media, each participant becomes part of the marketing department (or a challenge to the marketing effort) as other customers read their comments or reviews. The effectiveness of social media marketing campaigns is dependent on the promotion of online engagement. With the advent of social media marketing, it has become increasingly important to gain customer interest in products and services, which can eventually be translated into buying behavior, or voting and donating behavior in a political context. New online marketing concepts of engagement and loyalty have emerged which aim to build customer participation and brand reputation. Engagement in social media for the purpose of a social media strategy is divided into two parts. The first is proactive, regular posting of new online content, which can be seen through digital photos, digital videos, text, and conversations. It is also represented through sharing of content and information from others via weblinks. The second part is reactive conversations, with social media users responding to those who reach out to others' social media profiles through comments or messages. == Campaigns == === Local businesses === Small businesses use social networking sites as a promotional technique. Businesses can follow individuals' social media usage in their local area and advertise specials and deals, which can be exclusive and in the form of "get a free drink with a copy of this tweet". This type of message encourages other locals to follow the business on their official websites in order to obtain the promotional deal. The business's brand visibility is enhanced in the process. Social networking sites are also used by small businesses to develop their own market research on new products and services. By encouraging their customers to give feedback on new product ideas, businesses can gain insights on whether or not a product may be accepted by their target market enough to merit full production. In addition, customers will feel the company has engaged them in the process of co-creation—the process in which the business uses customer feedback to create or modify a product or service to fill a need of the target market. Such feedback can be presented in various forms, such as surveys, contests, and polls. Social networking sites such as LinkedIn, also provide opportunities for small businesses to find candidates to fill staff positions. Review sites such as Yelp help small businesses build their reputation beyond brand visibility. Positive customer peer reviews help influence new prospects to purchase goods and services more than company advertising. == Benefits == Social Media Marketing allows companies to promote themselves to large, diverse audiences that could not be reached through traditional marketing such as phone and email-based advertising. Marketing on most social media platforms also comes at little to no cost, making it accessible to virtually any size business. Social Media Marketing accommodates personalized and direct marketing that targets specific demographics and markets. Companies can engage with customers directly, allowing them to obtain feedback and resolve issues almost immediately. Another advantage of social media marketing is that it's an ideal environment for a company to conduct market research. It can be used
Social media use in hiring
Social media use in hiring refers to the examination by employers of job applicants' (public) social media profiles as part of the hiring assessment. For example, the vast majority of Fortune 500 companies use social media as a tool to screen prospective employees and as a tool for talent acquisition. This practice raises ethical questions. Employers and recruiters note that they have access only to information that applicants choose to make public. Many Western-European countries restrict employer's use of social media in the workplace. States including Arkansas, California, Colorado, Illinois, Maryland, Michigan, Nevada, New Jersey, New Mexico, Utah, Washington, and Wisconsin protect applicants and employees from surrendering usernames and passwords for social media accounts. Use of social media has caused significant problems for some applicants who are active on social media. A 2013 survey of 17,000 young people in six countries found that one in ten people aged 16 to 34 claimed to have been rejected for a job because of social media activity. Social media services have been reported to affect deception in resumes. While these services do not affect deception frequency, it does increase deception about interests and hobbies. == Ethical implications == This issue raises many ethical questions that some consider an employer's right and others consider discrimination. As of 2016, except in the states of California, Maryland, and Illinois, there are no laws that prohibit employers from using social media profiles as a basis of whether or not someone should be hired. Title VII also prohibits discrimination during any aspect of employment including hiring or firing, recruitment, or testing. Social media has been integrating into the workplace, and this has led to conflicts within employees and employers.[107] Particularly, Facebook has been seen as a popular platform for employers to investigate in order to learn more about potential employees. This conflict first started in Maryland when an employer requested and received an employee's Facebook username and password. State lawmakers first introduced legislation in 2012 to prohibit employers from requesting passwords to personal social accounts in order to get a job or to keep a job. This led to Canada, Germany, the U.S. Congress and 11 U.S. states to pass or propose legislation that prevents employers' access to private social accounts of employees.[108] Many Western European countries have already implemented laws that restrict the regulation of social media in the workplace. States including Arkansas, California, Colorado, Illinois, Maryland, Michigan, Nevada, New Jersey, New Mexico, Utah, Washington, and Wisconsin have passed legislation that protects potential employees and current employees from employers that demand them to give forth their username or password for a social media account. Laws that forbid employers from disciplining an employee based on activity off the job on social media sites have also been put into act in states including California, Colorado, Connecticut, North Dakota, and New York. Several states have similar laws that protect students in colleges and universities from having to grant access to their social media accounts. Eight states have passed the law that prohibits post secondary institutions from demanding social media login information from any prospective or current students and privacy legislation has been introduced or is pending in at least 36 states as of July 2013. As of May 2014, legislation has been introduced and is in the process of pending in at least 28 states and has been enacted in Maine and Wisconsin. In addition, the National Labor Relations Board has been devoting a lot of their attention to attacking employer policies regarding social media that can discipline employees who seek to speak and post freely on social media sites. Use of social media by young people has caused significant problems for some applicants who are active on social media when they try to enter the job market. A survey of 17,000 young people in six countries in 2013 found that 1 in 10 people aged 16 to 34 have been rejected for a job because of online comments they made on social media websites. A 2014 survey of recruiters found that 93% of them check candidates' social media postings. Moreover, professor Stijn Baert of Ghent University conducted a field experiment in which fictitious job candidates applied for real job vacancies in Belgium. They were identical except in one respect: their Facebook profile photos. It was found that candidates with the most wholesome photos were a lot more likely to receive invitations for job interviews than those with the more controversial photos. In addition, Facebook profile photos had a greater impact on hiring decisions when candidates were highly educated. These cases have created some privacy implications as to whether or not companies should have the right to look at employee's Facebook profiles. In March 2012, Facebook decided they might take legal action against employers for gaining access to employee's profiles through their passwords. According to Facebook Chief Privacy Officer for policy, Erin Egan, the company has worked hard to give its users the tools to control who sees their information. He also said users shouldn't be forced to share private information and communications just to get a job. According to the network's Statement of Rights and Responsibilities, sharing or soliciting a password is a violation of Facebook policy. Employees may still give their password information out to get a job, but according to Erin Egan, Facebook will continue to do their part to protect the privacy and security of their users. == Impacts == Use of social media by young people has caused significant problems for some applicants who are active on social media when they try to enter the job market. A survey of 17,000 young people in six countries in 2013 found that 1 in 10 people aged 16 to 34 have been rejected for a job because of online comments they made on social media websites. A 2014 survey of recruiters found that 93% of them check candidates' social media postings. Moreover, in 2015 professor Stijn Baert of Ghent University conducted a field experiment in which fictitious job candidates applied for real job vacancies in Belgium. They were identical except in one respect: their Facebook profile photos. It was found that candidates with the most wholesome photos were a lot more likely to receive invitations for job interviews than those with the more controversial photos. In addition, Facebook profile photos had a greater impact on hiring decisions when candidates were highly educated. These cases have created some privacy implications as to whether or not companies should have the right to look at employee's Facebook profiles. In March 2012, Facebook decided they might take legal action against employers for gaining access to employee's profiles through their passwords. According to Facebook Chief Privacy Officer for policy, Erin Egan, the company has worked hard to give its users the tools to control who sees their information. He also said users shouldn't be forced to share private information and communications just to get a job. According to the network's Statement of Rights and Responsibilities, sharing or soliciting a password is a violation of Facebook policy. Employees may still give their password information out to get a job, but according to Erin Egan, Facebook will continue to do their part to protect the privacy and security of their users. == Policy Responses == 26 US states now have laws against an employer requiring a current or potential employee to give the employer their username and password.
Data philanthropy
Data philanthropy refers to the practice of private companies donating corporate data. This data is usually donated to nonprofits or donation-run organizations that have difficulty keeping up with expensive data collection technology. The concept was introduced through the United Nations Global Pulse initiative in 2011 to explore corporate data assets for humanitarian, academic, and societal causes. For example, anonymized mobile data could be used to track disease outbreaks, or data on consumer actions may be shared with researchers to study public health and economic trends. == Definition == A large portion of data collected from the internet consists of user-generated content, such as blogs, social media posts, and information submitted through lead generation and data forms. Additionally, corporations gather and analyze consumer data to gain insight into customer behavior, identify potential markets, and inform investment decisions. United Nations Global Pulse director Robert Kirkpatrick has referred to this type of data as "massive passive data" or "data exhaust." == Challenges == While data philanthropy can enhance development policies, making users' private data available to various organizations raises concerns regarding privacy, ownership, and the equitable use of data. Different techniques, such as differential privacy and alphanumeric strings of information, can allow access to personal data while ensuring user anonymity. However, even if these algorithms work, re-identification may still be possible. Another challenge is convincing corporations to share their data. The data collected by corporations provides them with market competitiveness and insight regarding consumer behavior. Corporations may fear losing their competitive edge if they share the information they have collected with the public. Numerous moral challenges are also encountered. In 2016, Mariarosaria Taddeo, a digital ethics professor at the University of Oxford, proposed an ethical framework to address them. == Sharing strategies == The goal of data philanthropy is to create a global data commons where companies, governments, and individuals can contribute anonymous, aggregated datasets. The United Nations Global Pulse offers four different tactics that companies can use to share their data that preserve consumer anonymity: Share aggregated and derived data sets for analysis under nondisclosure agreements (NDA) Allow researchers to analyze data within the private company's own network under NDAs Real-Time Data Commons: data pooled and aggregated among multiple companies of the same industry to protect competitiveness Public/Private Alerting Network: companies mine data behind their own firewalls and share indicators == Application in various fields == Many corporations take part in data philanthropy, including social networking platforms (e.g., Facebook, Twitter), telecommunications providers (e.g., Verizon, AT&T), and search engines (e.g., Google, Bing). Collecting and sharing anonymized, aggregated user-generated data is made available through data-sharing systems to support research, policy development, and social impact initiatives. By participating in such efforts, these organizations contribute to causes regarded as beneficial to society, allowing institutions to give back meaningfully. With the onset of technological advancements, the sharing of data on a global scale and an in-depth analysis of these data structures could mitigate the effects of global issues such as natural disasters and epidemics. Robert Kirkpatrick, the Director of the United Nations Global Pulse, has argued that this aggregated information is beneficial for the common good and can lead to developments in research and data production in a range of varied fields. === Digital disease detection === Health researchers use digital disease detection by collecting data from various sources—such as social media platforms (e.g., Twitter, Facebook), mobile devices (e.g., cell phones, smartphones), online search queries, mobile apps, and sensor data from wearables and environmental sensors—to monitor and predict the spread of infectious diseases. This approach allows them to track and anticipate outbreaks of epidemics (e.g., COVID-19, Ebola), pandemics, vector-borne diseases (e.g., malaria, dengue fever), and respiratory illnesses (e.g., influenza, SARS), improving response and intervention strategies for the spread of diseases. In 2008, Centers for Disease Control and Prevention collaborated with Google and launched Google Flu Trends, a website that tracked flu-related searches and user locations to track the spread of the flu. Users could visit Google Flu Trends to compare the amount of flu-related search activity versus the reported numbers of flu outbreaks on a graphical map. One drawback of this method of tracking was that Google searches are sometimes performed due to curiosity rather than when an individual is suffering from the flu. According to Ashley Fowlkes, an epidemiologist in the CDC Influenza division, "The Google Flu Trends system tries to account for that type of media bias by modeling search terms over time to see which ones remain stable." Google Flu Trends is no longer publishing current flu estimates on the public website; however, visitors to the site can still view and download previous estimates. Current data can be shared with verified researchers. A study from the Harvard School of Public Health (HSPH), published in the October 12, 2012 issue of Science, discussed how phone data helped curb the spread of malaria in Kenya. The researchers mapped phone calls and texts made by 14,816,521 Kenyan mobile phone subscribers. When individuals left their primary living location, the destination and length of journey were calculated. This data was then compared to a 2009 malaria prevalence map to estimate the disease's commonality in each location. Combining all this information, the researchers could estimate the probability of an individual carrying malaria and map the movement of the disease. This research can be used to track the spread of similar diseases. === Humanitarian aid === Calling patterns of mobile phone users can determine the socioeconomic standings of the populace, which can be used to deduce "its access to housing, education, healthcare, and basic services such as water and electricity." Researchers from Columbia University and Karolinska Institute used daily SIM card location data from both before and after the 2010 Haiti earthquake to estimate the movement of people both in response to the earthquake and during the related 2010 Haiti cholera outbreak. Their research suggests that mobile phone data can provide rapid and accurate estimates of population movements during disasters and outbreaks of infectious disease. Big data can also provide information on looming disasters and can assist relief organizations in rapid-response and locating displaced individuals. By analyzing specific patterns within this 'big data', governments and NGOs can enhance responses to disruptive events such as natural disasters, disease outbreaks, and global economic crises. Leveraging real-time information enables a deeper understanding of individual well-being, allowing for more effective interventions. Corporations utilize digital services, such as human sensor systems, to detect and solve impending problems within communities. This is a strategy used by the private sector to anonymously share customer information for public benefit, while preserving user privacy. === Impoverished areas === Poverty still remains a worldwide issue, with over 2.5 billion people currently impoverished. Statistics indicate the widespread use of mobile phones, even within impoverished communities. Additional data can be collected through Internet access, social media, utility payments and governmental statistics. Data-driven activities can lead to the accumulation of 'big data', which in turn can assist international non-governmental organizations in documenting and evaluating the needs of underprivileged populations. Through data philanthropy, NGOs can distribute information while cooperating with governments and private companies. === Corporate === Data philanthropy incorporates aspects of social philanthropy by allowing corporations to create profound impacts through the act of giving back by dispersing proprietary datasets. The public sector collects and preserves information, considered an essential asset. Companies track and analyze users' online activities to gain insight into their needs related to new products and services. These companies view the welfare of the population as key to business expansion and progression by using their data to highlight global citizens' issues. Experts in the private sector emphasize the importance of integrating diverse data sources—such as retail, mobile, and social media data—to develop essential solutions for global challenges. In Data Philanthropy:
Prompt engineering
Prompt engineering is the process of structuring natural language inputs (known as prompts) to produce specified outputs from a generative artificial intelligence (GenAI) model. Context engineering is the related area of software engineering that focuses on the management of non-prompt contexts supplied to the GenAI model, such as metadata, API tools, and tokens. It can also be defined as the practice of designing and refining input instructions given to a generative AI model to produce more accurate, relevant, or useful outputs. Effective prompt engineering involves understanding how a model interprets language, and may include techniques such as few-shot prompting, chain-of-thought prompting, and role assignment. It is increasingly considered a skill for working with large language models (LLMs) in both research and professional contexts. During the 2020s AI boom, prompt engineering became regarded as a business capability across corporations and industries. Employees with the title prompt engineer were hired to create prompts that would increase productivity and efficacy, although the individual title has since lost traction amid AI models that produce better prompts than humans and corporate training in prompting for general employees. Common prompting techniques include multi-shot, chain-of-thought, and tree-of-thought prompting, as well as the use of assigning roles to the model. Automated prompt generation methods, such as retrieval-augmented generation (RAG), provide for greater accuracy and a wider scope of functions for prompt engineers. Prompt injection is a type of cybersecurity attack that targets machine learning models through malicious prompts. == Terminology == The Oxford English Dictionary defines prompt engineering as "The action or process of formulating and refining prompts for an artificial intelligence program, algorithm, etc., in order to optimize its output or to achieve a desired outcome; the discipline or profession concerned with this." In 2023, prompt ("an instruction given to an artificial intelligence program, algorithm, etc., which determines or influences the content it generates") was the runner-up to Oxford's word of the year. === Prompt === A prompt is some natural language text that describes and prescribes the task that an artificial intelligence (AI) should perform. A prompt for a text-to-text language model can be a query, a command, or a longer statement referencing context, instructions, and conversation history. The process of prompt engineering may involve designing clear queries, refining wording, providing relevant context, specifying the style of output, and assigning a character for the AI to mimic in order to guide the model toward more accurate, useful, and consistent responses. When communicating with a text-to-image or a text-to-audio model, a typical prompt contains a description of a desired output such as "a high-quality photo of an astronaut riding a horse" or "Lo-fi slow BPM electro chill with organic samples". Prompt engineering may be applied to text-to-image models to achieve a desired subject, style, layout, lighting, and aesthetic. === Techniques === Common terms used to describe various specific prompt engineering techniques include chain-of-thought, tree-of-thought, and retrieval-augmented generation (RAG). A 2024 survey of the field identified over 50 distinct text-based prompting techniques, 40 multimodal variants, and a vocabulary of 33 terms used across prompting research, highlighting a present lack of standardised terminology for prompt engineering. Vibe coding is an AI-assisted software development method where a user prompts an LLM with a description of what they want and lets it generate or edit the code. In 2025, "vibe coding" was the Collins Dictionary word of the year. === Context engineering === Context engineering is a related process that focuses on the context elements that accompany user prompts, which include system instructions, retrieved knowledge, tool definitions, conversation summaries, and task metadata. Context engineering is performed to improve reliability, provenance and token efficiency in production LLM systems. The concept emphasises operational practices such as token budgeting, provenance tags, versioning of context artifacts, observability (logging which context was supplied), and context regression tests to ensure that changes to supplied context do not silently alter system behaviour. == Rationale == Research has found that the performance of large language models (LLMs) is highly sensitive to choices such as the ordering of examples, the quality of demonstration labels, and even small variations in phrasing. In some cases, reordering examples in a prompt produced accuracy shifts of more than 40 percent. === In-context learning === A model's ability to temporarily learn from prompts is known as in-context learning. In-context learning is an emergent ability of large language models. It is an emergent property of model scale, meaning that breaks in scaling laws occur, leading to its efficacy increasing at a different rate in larger models than in smaller models. Unlike training and fine-tuning, which produce lasting changes, in-context learning is temporary. Training models to perform in-context learning can be viewed as a form of meta-learning, or "learning to learn". === Prompting to estimate model sensitivity === Research consistently demonstrates that LLMs are highly sensitive to subtle variations in prompt formatting, structure, and linguistic properties. Some studies have shown up to 76 accuracy points across formatting changes in few-shot settings. Linguistic features significantly influence prompt effectiveness—such as morphology, syntax, and lexico-semantic changes—which meaningfully enhance task performance across a variety of tasks. Clausal syntax, for example, improves consistency and reduces uncertainty in knowledge retrieval. This sensitivity persists even with larger model sizes, additional few-shot examples, or instruction tuning. To address sensitivity of models and make them more robust, several evaluative methods have been proposed. FormatSpread facilitates systematic analysis by evaluating a range of plausible prompt formats, offering a more comprehensive performance interval. Similarly, PromptEval estimates performance distributions across diverse prompts, enabling robust metrics such as performance quantiles and accurate evaluations under constrained budgets. == Prompting techniques == === Multi-shot === A prompt may include a few examples for a model to learn from in context, an approach called few-shot learning. For example, the prompt may ask the model to complete "maison → house, chat → cat, chien →", with the expected response being dog. === Chain-of-thought === Chain-of-thought (CoT) prompting is a technique that allows large language models (LLMs) to solve a problem as a series of intermediate steps before giving a final answer. In 2022, Google Brain reported that chain-of-thought prompting improves reasoning ability by inducing the model to answer a multi-step problem with steps of reasoning that mimic a train of thought. Chain-of-thought techniques were developed to help LLMs handle multi-step reasoning tasks, such as arithmetic or commonsense reasoning questions. When applied to PaLM, a 540 billion parameter language model, according to Google, CoT prompting significantly aided the model, allowing it to perform comparably with task-specific fine-tuned models on several tasks, achieving state-of-the-art results at the time on the GSM8K mathematical reasoning benchmark. It is possible to fine-tune models on CoT reasoning datasets to enhance this capability further and stimulate better interpretability. As originally proposed by Google, each CoT prompt is accompanied by a set of input/output examples—called exemplars—to demonstrate the desired model output, making it a few-shot prompting technique. However, according to a later paper from researchers at Google and the University of Tokyo, simply appending the words "Let's think step-by-step" was also effective, which allowed for CoT to be employed as a zero-shot technique. ==== Self-consistency ==== Self-consistency performs several chain-of-thought rollouts, then selects the most commonly reached conclusion out of all the rollouts. === Tree-of-thought === Tree-of-thought prompting generalizes chain-of-thought by generating multiple lines of reasoning in parallel, with the ability to backtrack or explore other paths. It can use tree search algorithms like breadth-first, depth-first, or beam. === Text-to-image prompting === In 2022, text-to-image models like DALL-E 2, Stable Diffusion, and Midjourney were released to the public. These models take text prompts as input and use them to generate images. Early text-to-image models typically do not understand negation, grammar and sentence structure in the same way as large language models, and may thus requi
IWARP
iWARP is a computer networking protocol that implements remote direct memory access (RDMA) for efficient data transfer over Internet Protocol networks. Contrary to some accounts, iWARP is not an acronym. Because iWARP is layered on Internet Engineering Task Force (IETF)-standard congestion-aware protocols such as Transmission Control Protocol (TCP) and Stream Control Transmission Protocol (SCTP), it makes few requirements on the network, and can be successfully deployed in a broad range of environments. == History == In 2007, the IETF published five Request for Comments (RFCs) that define iWARP: RFC 5040 A Remote Direct Memory Access Protocol Specification is layered over Direct Data Placement Protocol (DDP). It defines how RDMA Send, Read, and Write operations are encoded using DDP into headers on the network. RFC 5041 Direct Data Placement over Reliable Transports is layered over MPA/TCP or SCTP. It defines how received data can be directly placed into an upper layer protocols receive buffer without intermediate buffers. RFC 5042 Direct Data Placement Protocol (DDP) / Remote Direct Memory Access Protocol (RDMAP) Security analyzes security issues related to iWARP DDP and RDMAP protocol layers. RFC 5043 Stream Control Transmission Protocol (SCTP) Direct Data Placement (DDP) Adaptation defines an adaptation layer that enables DDP over SCTP. RFC 5044 Marker PDU Aligned Framing for TCP Specification defines an adaptation layer that enables preservation of DDP-level protocol record boundaries layered over the TCP reliable connected byte stream. These RFCs are based on the RDMA Consortium's specifications for RDMA over TCP. The RDMA Consortium's specifications are influenced by earlier RDMA standards, including Virtual Interface Architecture (VIA) and InfiniBand (IB). Since 2007, the IETF has published three additional RFCs that maintain and extend iWARP: RFC 6580 IANA Registries for the Remote Direct Data Placement (RDDP) Protocols published in 2012 defines IANA registries for Remote Direct Data Placement (RDDP) error codes, operation codes, and function codes. RFC 6581 Enhanced Remote Direct Memory Access (RDMA) Connection Establishment published in 2011 fixes shortcomings with iWARP connection setup. RFC 7306 Remote Direct Memory Access (RDMA) Protocol Extensions published in 2014 extends RFC 5040 with atomic operations and RDMA Write with Immediate Data. == Protocol == The main component in the iWARP protocol is the Direct Data Placement Protocol (DDP), which permits the actual zero-copy transmission. DDP itself does not perform the transmission; the underlying protocol (TCP or SCTP) does. However, TCP does not respect message boundaries; it sends data as a sequence of bytes without regard to protocol data units (PDU). In this regard, DDP itself may be better suited for SCTP, and indeed the IETF proposed a standard RDMA over SCTP. To run DDP over TCP requires a tweak known as marker PDU aligned (MPA) framing to guarantee boundaries of messages. Furthermore, DDP is not intended to be accessed directly. Instead, a separate RDMA protocol (RDMAP) provides the services to read and write data. Therefore, the entire RDMA over TCP specification is really RDMAP over DDP over either MPA/TCP or SCTP. All of these protocols can be implemented in hardware. Unlike IB, iWARP only has reliable connected communication, as this is the only service that TCP and SCTP provide. The iWARP specification omits other features of IB, such as Send with Immediate Data operations. With RFC 7306, the IETF is working to reduce these omissions. == Implementation == Because a kernel implementation of the TCP stack can be seen as a bottleneck, the protocol is typically implemented in hardware RDMA network interface controllers (rNICs). As simple data losses are rare in tightly coupled network environments, the error-correction mechanisms of TCP may be performed by software while the more frequently performed communications are handled strictly by logic embedded on the rNIC. Similarly, connections are often established entirely by software and then handed off to the hardware. Furthermore, the handling of iWARP specific protocol details is typically isolated from the TCP implementation, allowing rNICs to be used for both as RDMA offload and TCP offload (in support of traditional sockets based TCP/IP applications). The portion of the hardware implementation used for implementing the TCP protocol is known as the TCP Offload Engine (TOE). TOE itself does not prevent copying on the reception side, and must be combined with RDMA hardware for zero-copy results. The RDMA / TCP specification is a set of different wire protocols intended to be implemented in hardware (though it seems feasible to emulate it in software for compatibility but without the performance benefits). == Interfaces == iWARP is a protocol, not an implementation, but defines protocol behavior in terms of the operations that are legal for the protocol, known as Verbs. As such, iWARP does not have any single standard programming interface. However, programming interfaces tend to very closely correspond to the Verbs. Several programmatic interfaces have been proposed, including OpenFabrics Verbs, Network Direct, uDAPL, kDAPL, IT-API, and RNICPI. Implementations of some of these interfaces are available for different platforms, including Windows and Linux. == Services available == Networking services implemented over iWARP include those offered in the OpenFabrics Enterprise Distribution (OFED) by the OpenFabrics Alliance for Linux operating systems, and by Microsoft Windows via Network Direct. NVMe over Fabrics (NVMEoF) iSCSI Extensions for RDMA (iSER) Server Message Block Direct (SMB Direct) Sockets Direct Protocol (SDP) SCSI RDMA Protocol (SRP) Network File System over RDMA (NFS over RDMA) GPUDirect