As applied in the field of computer vision, graph cut optimization can be employed to efficiently solve a wide variety of low-level computer vision problems (early vision), such as image smoothing, the stereo correspondence problem, image segmentation, object co-segmentation, numerous military applications (eg Automatic target recognition) and many other problems that can be formulated in terms of energy minimization (eg Climate Science and Environmental modelling). Graph cut techniques are now increasingly being used in combination with more general spatial Artificial intelligence techniques (eg to enforce structure in Large language model output to sharpen tumour boundaries and similarly for various Augmented reality, Self-driving car, Robotics, Google Maps applications etc). Many of these energy minimization problems can be approximated by solving a maximum flow problem in a graph (and thus, by the max-flow min-cut theorem, define a minimal cut of the graph). Under most formulations of such problems in computer vision, the minimum energy solution corresponds to the maximum a posteriori estimate of a solution. Although many computer vision algorithms involve cutting a graph (e.g. normalized cuts), the term "graph cuts" is applied specifically to those models which employ a max-flow/min-cut optimization (other graph cutting algorithms may be considered as graph partitioning algorithms). "Binary" problems (such as denoising a binary image) can be solved exactly using this approach; problems where pixels can be labeled with more than two different labels (such as stereo correspondence, or denoising of a grayscale image) cannot be solved exactly, but solutions produced are usually near the global optimum. == History == The foundational theory of graph cuts in computer vision was first developed by Margaret Greig, Bruce Porteous and Allan Seheult (GPS) of Durham University in a now legendary discussion contribution to Julian Besag's 1986 paper and a more detailed follow on paper in 1989. In the Bayesian statistical context of smoothing noisy images, using a Markov random field as the image prior distribution, they showed with a mathematically beautiful proof how the maximum a posteriori estimate of a binary image can be obtained exactly by maximizing the flow through an associated image network, or graph, involving the introduction of a source and sink and Log-likelihood ratios. The problem was shown to be efficiently solvable exactly, an unexpected result as the problem was believed to be computationally intractable (NP hard). GPS also addressed the computational cost of the max-flow algorithm on large graphs, a significant concern at the time. They proposed a partitioning algorithm (see Section 4 of GPS) involving the recursive amalgamation of non-overlapping blocks, or tiles, which gave a 12X increase in speed. This approach recursively solved and amalgamated independent sub-graphs until the whole graph was solved. While contemporaries like Geman and Geman had advocated Parallel computing in the context of Simulated annealing, the GPS blocking strategy offered a deterministic structure amenable to parallelisation and anticipated modern artificial intelligence design across multiple GPUs. However, until recently, this aspect of the paper was largely ignored and subsequent research focused on Serial computer global search trees, such as the Boykov-Kolmogorov algorithm. Although the general k {\displaystyle k} -colour problem is NP hard for k > 2 , {\displaystyle k>2,} the GPS approach has turned out to have very wide applicability in general computer vision problems. This was first demonstrated by Boykov, Veksler and Zabih who, in a seminal paper published more than 10 years after the original GPS paper, and in other important works, lit the blue touch paper for the general adoption of graph cut techniques in computer vision. They showed that, for general problems, the GPS approach can be applied iteratively to sequences of binary problems, using their now ubiquitous alpha-expansion algorithm, yielding near optimal solutions. Prior to these results, approximate local optimisation techniques such as simulated annealing (as proposed by the Geman brothers) or iterated conditional modes (a type of greedy algorithm suggested by Julian Besag) were used to solve such image smoothing problems. Building on these advancements, GPS graph cut optimization was subsequently adapted for interactive image segmentation, most notably through the "GrabCut" algorithm introduced by Carsten Rother, Vladimir Kolmogorov, and Andrew Blake of Microsoft Research, Cambridge. GrabCut extended earlier interactive graph cut methods by replacing monochrome image histograms with Gaussian mixture models to estimate colour distributions, and by employing an iterative GPS energy minimisation scheme. This approach significantly simplified user interaction, requiring only a rough bounding box around the target object rather than detailed user-drawn strokes, and it quickly became a standard tool in both academic research and commercial image editing software. The GPS paper connected and bridged profound ideas from Mathematical statistics (Bayes' theorem, Markov random field), Physics (Ising model), Optimisation (Energy function) and Computer science (Network flow problem) and led the move away from approximate local and slow optimisation approaches (eg simulated annealing) to more powerful exact, or near exact, faster global optimisation techniques. It is now recognised as seminal as it was well ahead of its time and, in particular, was published years before the computing power revolution of Moore's law and GPUs. Significantly, GPS was published in a mathematical statistics (rather than a computer vision) journal, and this led to it being overlooked by the computer vision community for many years. It is unofficially known as "The Velvet Underground" paper of computer vision (ie although very few computer vision people read the paper [bought the record], those that did, most importantly Boykov, Veksler and Zabih, started new and important research [formed a band]). This is confirmed by GPS' very large amplification ratio (2nd order citations/first order citations), estimated at well in excess of 100. Despite the foundational nature of the GPS work, formal recognition from the computer vision community has predominantly gone to the researchers who followed to extend and popularise the graph cut method. For example, Boykov, Veksler and Zabih deservedly received a Helmholtz Prize from the ICCV in 2011. This prize recognises ICCV papers from 10 or more years earlier that have had a significant impact on computer vision research. In 2011, Couprie et al. proposed a general image segmentation framework, called the "Power Watershed", that minimized a real-valued indicator function from [0,1] over a graph, constrained by user seeds (or unary terms) set to 0 or 1, in which the minimization of the indicator function over the graph is optimized with respect to an exponent p {\displaystyle p} . When p = 1 {\displaystyle p=1} , the Power Watershed is optimized by graph cuts, when p = 0 {\displaystyle p=0} the Power Watershed is optimized by shortest paths, p = 2 {\displaystyle p=2} is optimized by the random walker algorithm and p = ∞ {\displaystyle p=\infty } is optimized by the watershed algorithm. In this way, the Power Watershed may be viewed as a generalization of graph cuts that provides a straightforward connection with other energy optimization segmentation/clustering algorithms. == Binary segmentation of images == === Notation === Image: x ∈ { R , G , B } N {\displaystyle x\in \{R,G,B\}^{N}} Output: Segmentation (also called opacity) S ∈ R N {\displaystyle S\in R^{N}} (soft segmentation). For hard segmentation S ∈ { 0 for background , 1 for foreground/object to be detected } N {\displaystyle S\in \{0{\text{ for background}},1{\text{ for foreground/object to be detected}}\}^{N}} Energy function: E ( x , S , C , λ ) {\displaystyle E(x,S,C,\lambda )} where C is the color parameter and λ is the coherence parameter. E ( x , S , C , λ ) = E c o l o r + E c o h e r e n c e {\displaystyle E(x,S,C,\lambda )=E_{\rm {color}}+E_{\rm {coherence}}} Optimization: The segmentation can be estimated as a global minimum over S: arg min S E ( x , S , C , λ ) {\displaystyle {\arg \min }_{S}E(x,S,C,\lambda )} === Existing methods === Standard Graph cuts: optimize energy function over the segmentation (unknown S value). Iterated Graph cuts: First step optimizes over the color parameters using K-means. Second step performs the usual graph cuts algorithm. These 2 steps are repeated recursively until convergence Dynamic graph cuts:Allows to re-run the algorithm much faster after modifying the problem (e.g. after new seeds have been added by a user). === Energy function === Pr ( x ∣ S ) = K − E {\displaystyle \Pr(x\mid S)=K^{-E}} where the energy E {\displaystyle E} is composed of two different mod
Traceability
Traceability is the capability to trace something. In some cases, it is interpreted as the ability to verify the history, location, or application of an item by means of documented recorded identification. Other common definitions include the capability (and implementation) of keeping track of a given set or type of information to a given degree, or the ability to chronologically interrelate uniquely identifiable entities in a way that is verifiable. Traceability is applicable to measurement, supply chain, software development, healthcare and security. == Measurement == The term measurement traceability or metrological traceability is used to refer to an unbroken chain of comparisons relating an instrument's measurements to a known standard. Calibration to a traceable standard can be used to determine an instrument's bias, precision, and accuracy. It may also be used to show a chain of custody—from current interpretation of evidence to the actual evidence in a legal context, or history of handling of any information. In many countries, national standards for weights and measures are maintained by a National Metrological Institute (NMI) which provides the highest level of standards for the calibration / measurement traceability infrastructure in that country. Examples of government agencies include the National Physical Laboratory, UK (NPL) the National Institute of Standards and Technology (NIST) in the USA, the Physikalisch-Technische Bundesanstalt (PTB) in Germany, the Instituto Nazionale di Ricerca Metrologica (INRiM) in Italy, and the National Research Council of Canada (NRC). As defined by NIST, "Traceability of measurement requires the establishment of an unbroken chain of comparisons to stated references each with a stated uncertainty." A clock providing traceable time is traceable to a time standard such as Coordinated Universal Time or International Atomic Time. The Global Positioning System is a source of traceable time. === Food processing === In food processing (meat processing, fresh produce processing), the term traceability refers to the recording through means of barcodes or RFID tags and other tracking media, all movement of product and steps within the production process. One of the key reasons this is such a critical point is in instances where an issue of contamination arises, and a recall is required. Where traceability has been closely adhered to, it is possible to identify, by precise date/time and exact location which goods must be recalled, and which are safe, potentially saving millions of dollars in the recall process. Traceability within the food processing industry is also utilised to identify key high production and quality areas of a business, versus those of low return, and where points in the production process may be improved. In food processing software, traceability systems imply the use of a unique piece of data (e.g., order date/time or a serialized sequence number, generally through the use of a barcode / RFID) which can be traced through the entire production flow, linking all sections of the business, including suppliers and future sales through the supply chain. Messages and files at any point in the system can then be audited for correctness and completeness, using the traceability software to find the particular transaction and/or product within the supply chain. In food systems, ISO 22005, as part of the ISO 22000 family of standards, has been developed to define the principles for food traceability and specifies the basic requirements for the design and implementation of a feed and food traceability system. It can be applied by an organization operating at any step in the feed and food chain. The European Union's General Food Law came into force in 2002, making traceability compulsory for food and feed operators and requiring those businesses to implement traceability systems. The EU introduced its Trade Control and Expert System, or TRACES, in April 2004. The system provides a central database to track movement of animals within the EU and from third countries. Australia has its National Livestock Identification System to keep track of livestock from birth to slaughterhouse. India has started taking initiatives for setting up traceability systems at Government and Corporate levels. Grapenet, an initiative by Agriculture and Processed Food Products Export Development Authority (APEDA), Ministry of Commerce, Government of India is an example in this direction. GrapeNet is an internet based traceability software system for monitoring fresh grapes exported from India to the European Union. GrapeNet is a first of its kind initiative in India that has put in place an end-to-end system for monitoring pesticide residue, achieve product standardization and facilitate tracing back from pallets to the farm of the Indian grower, through the various stages of sampling, testing, certification and packing. Grapenet won the National Award (Gold), in the winners announced for the best e-Governance initiatives undertaken in India in 2007. The Directorate Generate Foreign Trade (DGFT), Government of India, through its notification dated 04.02.2009 relating to Amendment in Foreign Trade Policy (RE2008)has mandated that Export to the European Union is permitted subject to registration with APEDA, thereby making Grapenet mandatory for all exports of fresh grapes from India to Europe. Uruguay has also designed a system called "Traceability & Electronic Information System of the Beef Industry". Traceability in food supply can also refer to practices employed by individual companies, including Ritual and Amway's Nutrilite. In the case of Nutrilite's supplements, ingredients are documented and tested throughout farming, processing, and manufacturing to ensure traceability at each stage of production. == Systems and software development == In systems and software development, the term traceability (or requirements traceability) refers to the ability to link product requirements back to stakeholders' rationales and forward to corresponding design artifacts, code, and test cases. Traceability supports numerous software engineering activities such as change impact analysis, compliance verification or traceback of code, regression test selection, and requirements validation. It is usually accomplished in the form of a matrix created for the verification and validation of the project. Unfortunately, the practice of constructing and maintaining a requirements trace matrix (RTM) can be very arduous and over time the traces tend to erode into an inaccurate state unless date/time stamped. Alternate automated approaches for generating traces using information retrieval methods have been developed. The IEEE defines traceability as "(1)The degree to which a relationship can be established between two or more products of the development process, especially products having a predecessor, successor or master-subordinate relationship to one another. For example, the degree to which the requirements and design of a given software component match. See also: consistency. " and "(2) The degree to which each element in a software development product establishes its reason for existing; for example, the degree to which each element in a bubble chart references the requirement that it satisfies." In transaction processing software, traceability implies use of a unique piece of data (e.g., order date/time or a serialized sequence number) which can be traced through the entire software flow of all relevant application programs. Messages and files at any point in the system can then be audited for correctness and completeness, using the traceability key to find the particular transaction. This is also sometimes referred to as the transaction footprint. == Health care == Patient safety during healthcare service plays an important role in preventing delayed recovery or even mortality, by increasing and improving the quality of life of citizens, and is considered an indicator of the quality status of health services Maintaining patient safety is a complex task and involves factors inherent to the environment and human actions. New technologies facilitate the traceability tools of patients and medications. This is particularly relevant for drugs that are considered high risk and cost. Recent research in the healthcare industry emphasizes the significant impact of Blockchain Technology (BCT) on improving the performance of healthcare supply chain management. It highlights BCT's role in enhancing transparency, data immutability, and efficient management, leading to better cooperation among stakeholders and effective risk mitigation in healthcare services. The World Health Organization has recognized the importance of traceability for medical products of human origin (MPHO) and urged member states "to encourage the implementation of globally consistent coding systems to facilitate national and international traceability". == Security and cri
Human-in-the-loop
Human-in-the-loop (HITL) is used in multiple contexts. It can be defined as a model requiring human interaction. HITL is associated with modeling and simulation (M&S) in the live, virtual, and constructive taxonomy. HITL, along with the related human-on-the-loop, are also used in relation to lethal autonomous weapons. Further, HITL is used in the context of machine learning.It is also used in conversational AI to manage complex interactions that require human empathy. == Machine learning == In machine learning, HITL is used in the sense of humans aiding the computer in making the correct decisions in building a model. HITL improves machine learning over random sampling by selecting the most critical data needed to refine the model. == Simulation == In simulation, HITL models may conform to human factors requirements as in the case of a mockup. In this type of simulation, a human is always part of the simulation and consequently influences the outcome in such a way that is difficult if not impossible to reproduce exactly. HITL also readily allows for the identification of problems and requirements that may not be easily identified by other means of simulation. HITL is often referred to as an interactive simulation, which is a special kind of physical simulation in which physical simulations include human operators, such as in a flight or a driving simulator. === Benefits === Human-in-the-loop allows the user to change the outcome of an event or process. The immersion effectively contributes to a positive transfer of acquired skills into the real world. This can be demonstrated by trainees utilizing flight simulators in preparation to become pilots. HITL also allows for the acquisition of knowledge regarding how a new process may affect a particular event. Utilizing HITL allows participants to interact with realistic models and attempt to perform as they would in an actual scenario. HITL simulations bring to the surface issues that would not otherwise be apparent until after a new process has been deployed. A real-world example of HITL simulation as an evaluation tool is its usage by the Federal Aviation Administration (FAA) to allow air traffic controllers to test new automation procedures by directing the activities of simulated air traffic while monitoring the effect of the newly implemented procedures. As with most processes, there is always the possibility of human error, which can only be reproduced using HITL simulation. Although much can be done to automate systems, humans typically still need to take the information provided by a system to determine the next course of action based on their judgment and experience. Intelligent systems can only go so far in certain circumstances to automate a process; only humans in the simulation can accurately judge the final design. Tabletop simulation may be useful in the very early stages of project development for the purpose of collecting data to set broad parameters, but the important decisions require human-in-the-loop simulation. HITL reflects scenarios where human input remains essential despite advances in automation. === Within the virtual simulation taxonomy === Virtual simulations inject HITL in a central role by exercising motor control skills (e.g. flying an airplane), decision making skills (e.g. committing fire control resources to action), or communication skills (e.g. as members of a C4I team). === Examples === Flight simulators Driving simulators Marine simulators Video games Supply chain management simulators Digital puppetry === Misconceptions === Although human-in-the-loop simulation can include a computer simulation in the form of a synthetic environment, computer simulation is not necessarily a form of human-in-the-loop simulation, and is often considered as human-out-of-the loop simulation. In this particular case, a computer model’s behavior is modified according to a set of initial parameters. The results of the model differ from the results stemming from a true human-in-the-loop simulation because the results can easily be replicated time and time again, by simply providing identical parameters. == Weapons == === Taxonomy === Three classifications of the degree of human control of autonomous weapon systems were laid out by Bonnie Docherty in a 2012 Human Rights Watch report. human-in-the-loop: a human must instigate the action of the weapon (in other words not fully autonomous) human-on-the-loop: a human may abort an action human-out-of-the-loop: no human action is involved === Positive human action === In discussions of autonomous weapons and nuclear command and control, the phrase positive human action has been used alongside "human-in-the-loop" to emphasize that a human operator must affirmatively authorize the use of force. Descriptions of the United States Navy's Aegis Combat System have used the phrase in characterizing a requirement for affirmative human action to initiate live firing. A survey of autonomous weapons systems described the Aegis "Auto SM" mode as one in which "the system fully develops the engagement process however engagement requires positive human action". The phrase entered United States federal law in the National Defense Authorization Act for Fiscal Year 2025, which stipulates that artificial intelligence systems not compromise "the principle of requiring positive human actions in execution of decisions by the President with respect to the employment of nuclear weapons".
And–or tree
An and–or tree is a graphical representation of the reduction of problems (or goals) to conjunctions and disjunctions of subproblems (or subgoals). == Example == The and–or tree: represents the search space for solving the problem P, using the goal-reduction methods: P if Q and R P if S Q if T Q if U == Definitions == Given an initial problem P0 and set of problem solving methods of the form: P if P1 and … and Pn the associated and–or tree is a set of labelled nodes such that: The root of the tree is a node labelled by P0. For every node N labelled by a problem or sub-problem P and for every method of the form P if P1 and ... and Pn, there exists a set of children nodes N1, ..., Nn of the node N, such that each node Ni is labelled by Pi. The nodes are conjoined by an arc, to distinguish them from children of N that might be associated with other methods. A node N, labelled by a problem P, is a success node if there is a method of the form P if nothing (i.e., P is a "fact"). The node is a failure node if there is no method for solving P. If all of the children of a node N, conjoined by the same arc, are success nodes, then the node N is also a success node. Otherwise the node is a failure node. == Search strategies == An and–or tree specifies only the search space for solving a problem. Different search strategies for searching the space are possible. These include searching the tree depth-first, breadth-first, or best-first using some measure of desirability of solutions. The search strategy can be sequential, searching or generating one node at a time, or parallel, searching or generating several nodes in parallel. == Relationship with logic programming == The methods used for generating and–or trees are propositional logic programs (without variables). In the case of logic programs containing variables, the solutions of conjoint sub-problems must be compatible. Subject to this complication, sequential and parallel search strategies for and–or trees provide a computational model for executing logic programs. == Relationship with two-player games == And–or trees can also be used to represent the search spaces for two-person games. The root node of such a tree represents the problem of one of the players winning the game, starting from the initial state of the game. Given a node N, labelled by the problem P of the player winning the game from a particular state of play, there exists a single set of conjoint children nodes, corresponding to all of the opponents responding moves. For each of these children nodes, there exists a set of non-conjoint children nodes, corresponding to all of the player's defending moves. For solving game trees with proof-number search family of algorithms, game trees are to be mapped to and–or trees. MAX-nodes (i.e. maximizing player to move) are represented as OR nodes, MIN-nodes map to AND nodes. The mapping is possible, when the search is done with only a binary goal, which usually is "player to move wins the game".
Reasoning model
A reasoning model, also known as a reasoning language model (RLM) or large reasoning model (LRM), is a type of large language model (LLM) that has been specifically trained to solve complex tasks requiring multiple steps of logical reasoning. These models demonstrate superior performance on logic, mathematics, and programming tasks compared to standard LLMs. They possess the ability to revisit and revise earlier reasoning steps and utilize additional computation during inference as a method to scale performance, complementing traditional scaling approaches based on training data size, model parameters, and training compute. == Overview == Unlike traditional language models that generate responses immediately, reasoning models allocate additional compute, or thinking, time before producing an answer to solve multi-step problems. OpenAI introduced this terminology in September 2024 when it released the o1 series, describing the models as designed to "spend more time thinking" before responding. The company framed o1 as a reset in model naming that targets complex tasks in science, coding, and mathematics, and it contrasted o1's performance with GPT-4o on benchmarks such as AIME and Codeforces. Independent reporting the same week summarized the launch and highlighted OpenAI's claim that o1 automates chain-of-thought style reasoning to achieve large gains on difficult exams. In operation, reasoning models generate internal chains of intermediate steps, then select and refine a final answer. OpenAI reported that o1's accuracy improves as the model is given more reinforcement learning during training and more test-time compute at inference. The company initially chose to hide raw chains and instead return a model-written summary, stating that it "decided not to show" the underlying thoughts so researchers could monitor them without exposing unaligned content to end users. Commercial deployments document separate "reasoning tokens" that meter hidden thinking and a control for "reasoning effort" that tunes how much compute the model uses. These features make the models slower than ordinary chat systems while enabling stronger performance on difficult problems. == History == The research trajectory toward reasoning models combined advances in supervision, prompting, and search-style inference. Early alignment work on reinforcement learning from human feedback showed that models can be fine-tuned to follow instructions with "human feedback" and preference-based rewards. In 2022, Google Research scientists Jason Wei and Denny Zhou showed that chain-of-thought prompting "significantly improves the ability" of large models on complex reasoning tasks. Input → Step 1 → Step 2 → ⋯ → Step n ⏟ Reasoning chain → Answer {\displaystyle {\text{Input}}\rightarrow \underbrace {{\text{Step}}_{1}\rightarrow {\text{Step}}_{2}\rightarrow \cdots \rightarrow {\text{Step}}_{n}} _{\text{Reasoning chain}}\rightarrow {\text{Answer}}} A companion result demonstrated that the simple instruction "Let's think step by step" can elicit zero-shot reasoning. Follow-up work introduced self-consistency decoding, which "boosts the performance" of chain-of-thought by sampling diverse solution paths and choosing the consensus, and tool-augmented methods such as ReAct, a portmanteau of Reason and Act, that prompt models to "generate both reasoning traces" and actions. Research then generalized chain-of-thought into search over multiple candidate plans. The Tree-of-Thoughts framework from Princeton computer scientist Shunyu Yao proposes that models "perform deliberate decision making" by exploring and backtracking over a tree of intermediate thoughts. OpenAI's reported breakthrough focused on supervising reasoning processes rather than only outcomes, with Lightman et al.'s "Let's Verify Step by Step" reporting that rewarding each correct step "significantly outperforms outcome supervision" on challenging math problems and improves interpretability by aligning the chain-of-thought with human judgment. OpenAI's o1 announcement ties these strands together with a large-scale reinforcement learning algorithm that trains the model to refine its own chain of thought, and it reports that accuracy rises with more training compute and more time spent thinking at inference. Together, these developments define the core of reasoning models. They use supervision signals that evaluate the quality of intermediate steps, they exploit inference-time exploration such as consensus or tree search, and they expose controls for how much internal thinking compute to allocate. OpenAI's o1 family made this approach available at scale in September 2024 and popularized the label "reasoning model" for LLMs that deliberately think before they answer. The development of reasoning models illustrates Richard S. Sutton's "bitter lesson" that scaling compute typically outperforms methods based on human-designed insights. This principle was demonstrated by researchers at the Generative AI Research Lab (GAIR), who initially attempted to replicate o1's capabilities using sophisticated methods including tree search and reinforcement learning in late 2024. Their findings, published in the "o1 Replication Journey" series, revealed that knowledge distillation, a comparatively straightforward technique that trains a smaller model to mimic o1's outputs, produced unexpectedly strong performance. This outcome illustrated how direct scaling approaches can, at times, outperform more complex engineering solutions. === Drawbacks === Reasoning models require significantly more computational resources during inference compared to non-reasoning models. Research on the American Invitational Mathematics Examination (AIME) benchmark found that reasoning models were 10 to 74 times more expensive to operate than their non-reasoning counterparts. The extended inference time is attributed to the detailed, step-by-step reasoning outputs that these models generate, which are typically much longer than responses from standard large language models that provide direct answers without showing their reasoning process. One researcher in early 2025 argued that these models may face potential additional denial-of-service concerns with "overthinking attacks." === Releases === ==== 2024 ==== In September 2024, OpenAI released o1-preview, a large language model with enhanced reasoning capabilities. The full version, o1, was released in December 2024. OpenAI initially shared preliminary results on its successor model, o3, in December 2024, with the full o3 model becoming available in 2025. Alibaba released reasoning versions of its Qwen large language models in November 2024. In December 2024, the company introduced QvQ-72B-Preview, an experimental visual reasoning model. In December 2024, Google introduced Deep Research in Gemini, a feature designed to conduct multi-step research tasks. On December 16, 2024, researchers demonstrated that by scaling test-time compute, a relatively small Llama 3B model could outperform a much larger Llama 70B model on challenging reasoning tasks. This experiment suggested that improved inference strategies can unlock reasoning capabilities even in smaller models. ==== 2025 ==== In January 2025, DeepSeek released R1, a reasoning model that achieved performance comparable to OpenAI's o1 at significantly lower computational cost. The release demonstrated the effectiveness of Group Relative Policy Optimization (GRPO), a reinforcement learning technique used to train the model. On January 25, 2025, DeepSeek enhanced R1 with web search capabilities, allowing the model to retrieve information from the internet while performing reasoning tasks. Research during this period further validated the effectiveness of knowledge distillation for creating reasoning models. The s1-32B model achieved strong performance through budget forcing and scaling methods, reinforcing findings that simpler training approaches can be highly effective for reasoning capabilities. On February 2, 2025, OpenAI released Deep Research, a feature powered by their o3 model that enables users to conduct comprehensive research tasks. The system generates detailed reports by automatically gathering and synthesizing information from multiple web sources. OpenAI called GPT-4.5 its "last non-chain-of-thought model", and implemented with GPT-5 a router model that selects a model based on the difficulty of the task. ==== 2026 ==== In January 2026, Moonshot AI released Kimi K2.5, an open-source 1 trillion parameter MoE model with 32 billion active parameters. It uses an “Agent Swarm” system that dynamically decomposes tasks into sub-agents for reasoning and execution, enabling more scalable multi-step problem solving than a single sequential reasoning chain. == Training == Reasoning models follow the familiar large-scale pretraining used for frontier language models, then diverge in the post-training and optimization. OpenAI reports that o1 is trained with a large-
Microsoft Sway
Microsoft Sway is a presentation program and is part of the Microsoft 365 family of products. Sway was offered for general release by Microsoft in August 2015. It allows users who have a Microsoft account to combine text and media to create a presentable website. Users can pull content locally from the device in use, or from internet sources such as Bing, Facebook, OneDrive, and YouTube. Sway is distinguished from Microsoft FrontPage and Microsoft Expression Web – unrelated web design programs previously developed by Microsoft – in that Sway includes a method for hosting sites. Sway sites are stored on Microsoft's servers and are tied to the user's Microsoft account. They can be viewed and edited from any web browser through Office on the web. There is no offline editing or viewing function, but sites can be accessed using the app for Windows, and formerly iOS. == History == Sway was developed internally by Microsoft. In late 2014, the company announced an invite-only preview version of Sway and announced that Sway would not require an Office 365 subscription. An iOS app was released as a preview on 31 October 2014, but was discontinued on 17 December 2018 due to low usage. As of July 17, 2021, the Sway iOS app's discontinuance in 2018 was the last piece of news posted in the Sway tech blog. The Sway feature blog has not received an update since April 2017. The Microsoft Office Roadmap did not include any items related to Sway ever since. The iOS application is no longer under active development, and is not available for download. Since 2023, Microsoft has been consolidating the domains of its Microsoft 365 apps and services under cloud.microsoft. By 2025, the vast majority of services, including Sway, have already migrated to the cloud.microsoft domain. == Features == Users are able to add content from various sources into their Sway presentations. Some of the integrated services are owned by Microsoft, including OneNote, Bing, and other Sway sites. The program also provides native integration with other services, including YouTube, Facebook, Twitter, Mixcloud, and Infogram.
Perusall
Perusall is a social web annotation tool intended for use by students at schools and universities. It allows users to annotate the margins of a text in a virtual group setting that is similar to social media—with upvoting, emojis, chat functionality, and notification. It also includes automatic AI grading. == History == Perusall began as a research project at Harvard University. It later became an educational product for students and teachers. As of 2024, Perusall states more than 5 million students have used the tool at over 5,000 educational institutions in 112 countries." == Functionality == Perusall integrates with learning management systems such as Moodle, Canvas and Blackboard to aid with collaborative annotation. The tool supports annotation of a range of media including text, images, equations, videos, PDFs and snapshots of webpages.