25 skills found
EveryInc / Proof SDKProof SDK: open-source collaborative editor, provenance model, and agent HTTP bridge
trungdong / ProvA Python library for W3C Provenance Data Model (PROV)
lucmoreau / ProvToolboxJava toolkit to create and convert W3C PROV data model representations, and build provenance-enabled applications in a variety of programming languages (java, python, typescript, javascript)
devallibus / ShiplogSUPERCHARGE AI-assisted development by using Git. Cross-model review gates, evidence-linked closure, verification profiles, model-tier routing, artifact envelopes, and provenance signing — all from a single skill for Claude Code, Codex, and Cursor.
incf-nidash / Nidm SpecsNeuroimaging Data Model (NIDM): describing neuroimaging data and provenance
lucmoreau / OpenProvenanceModelThe overarching project of Java code related to the Open Provenance specifications.
YinBo0927 / ReProThe official code of Refinement Provenance Inference: Detecting LLM-Refined Training Prompts from Model Behavior
IntelLabs / Atlas CliA command-line interface tool for creating, managing, and verifying Content Provenance and Authenticity (C2PA) manifests for machine learning models, datasets, and related artifacts.
tongyu0924 / Secure Diffusion Watermarking SurveyA curated collection of papers on watermarking, attribution, and provenance in diffusion models. Based on the Secure Diffusion survey, this repo organizes key techniques and resources for safeguarding generative content.
andrewmogbolu2 / Blockchain TechnologyBlockchain and AI are on just about every chief information officers watchlist of game-changing technologies that stand to reshape industries. Both technologies come with immense benefits, but both also bring their own challenges for adoption. It is also fair to say that the hype surrounding these technologies individually may be unprecedented, so the thought of bringing these two ingredients together may be viewed by some as brewing a modern-day version of IT pixie dust. At the same time, there is a logical way to think about this mash-up that is both sensible and pragmatic. Today, AI is for all intents and purposes a centralized process. An end user must have extreme faith in the central authority to produce a trusted business outcome. By decentralizing the three key elements of AI — that is, data, models, and analytics — blockchain can deliver the trust and confidence often needed for end users to fully adopt and rely on AI-based business processes. Let’s explore how blockchain is poised to enrich AI by bringing trust to data, models and analytics. Your data is your data Many of the world’s most notable AI technology services are centralized — including Amazon, Apple, Facebook, Google, as well as Chinese companies Alibaba, Baidu and Tencent. Yet all have encountered challenges in establishing trust among their eager, but somewhat cautious users. How can a business provide assurance to its users that its AI has not overstepped its bounds? Imagine if these AI services could produce a “forensic report,” verified by a third party, to prove to you, beyond a reasonable doubt, how and when businesses are using your data once those are ingested. Imagine further that your data could be used only if you gave permission to do so. A blockchain ledger can be used as a digital rights management system, allowing your data to be “licensed” to the AI provider under your terms, conditions and duration. The ledger would act as an access management system storing the proofs and permission by which a business can access and use the user’s data. Trusted AI models Consider the example of using blockchain technology as a means of providing trusted data and provenance of training models for machine learning. In this case, we’ve created a fictitious system to answer the question of whether a fruit is an apple or orange. This question-answering system that we build is called a model, and this model is created via a process called training. The goal of training is to create an accurate model that answers our questions correctly most of the time. Of course, to train a model, we need to collect data to train on — for this example, that could be the color of the fruit (as a wavelength of light) and the sugar content (as a percentage). With blockchain, you can track the provenance of the training data as well as see an audit trail of the evidence that led to the prediction of why a particular fruit is considered an apple versus an orange. A business can also prove that it is not “juicing up” its books by tagging fruit more often as apples, if that is the more expensive of the two fruits. Explaining AI decisions The European Union has adopted a law requiring that any decision made by a machine be readily explainable, on penalty of fines that could cost companies billions of dollars. The EU General Data Protection Regulation (GDPR), which came into force in 2018, includes a right to obtain an explanation of decisions made by algorithms and a right to opt out of some algorithmic decisions altogether. Massive amounts of data are being produced every second — more data than humans have the ability to assess and use as the basis for drawing conclusions. However, AI applications are capable of assessing large data sets and many variables, while learning about or connecting those variables relevant to its tasks and objectives. For this very reason, AI continues to be adopted in various industries and applications, and we are relying more and more on their outcomes. It is essential, however, that any decisions made by AI are still verified for accuracy by humans. Blockchain can help clarify the provenance, transparency, understanding, and explanations of those outcomes and decisions. If decisions and associated data points are recorded via transactions on a blockchain, the inherent attributes of blockchain will make auditing them much simpler. Blockchain is a key technology that brings trust to transactions in a network; therefore, infusing blockchain into AI decision-making processes could be the element needed to achieve the transparency necessary to fully trust the decisions and outcomes derived from AI. Blockchain and the Internet of Things More than a billion intelligent, connected devices are already part of today’s IoT. The expected proliferation of hundreds of billions more places us at the threshold of a transformation sweeping across the electronics industry and many other areas. With the advancement in IoT, industries are now enabled to capture data, gain insight from the data, and make decisions based on the data. Therefore, there is a lot of “trust” in the information obtained. But the real truth of the matter is, do we really know where these data came from and should we be making decisions and transacting based on data we cannot validate? For example, did weather data really originate from a censor in the Atlantic Ocean or did the shipping container really not exceed the agreed temperature limit? The IoT use cases are massive, but they all share the same issue with trust. IoT with blockchain can bring real trust to captured data. The underlying idea is to give devices, at the time of their creation, an identity that can be validated and verified throughout their lifecycle with blockchain. There is great potential for IoT systems in blockchain technology capabilities that rely on device identity protocols and reputation systems. With a device identity protocol, each device can have its own blockchain public key and send encrypted challenge and response messages to other devices, thereby ensuring a device remains in control of its identity. In addition, a device with an identity can develop a reputation or history that is tracked by a blockchain. Smart contracts represent the business logic of a blockchain network. When a transaction is proposed, these smart contracts are autonomously executed within the guidelines set by the network. In IoT networks, smart contracts can play a pivotal role by providing automated coordination and authorization for transactions and interactions. The original idea behind IoT was to surface data and gain actionable insight at the right time. For example, smart homes are a thing of the present and most everything can be connected. In fact, with IoT, when something goes wrong, these IoT devices can even take action — for example, ordering a new part. We need a way to govern the actions taken by these devices, and smart contracts are a great way to do so. In an ongoing experiment I have followed in Brooklyn, New York, a community is using a blockchain to record the production of solar energy and enable the purchase of excess renewable energy credits. The device itself has an identity and builds a reputation through its history of records and exchange. Through the blockchain, people can aggregate their purchasing power more easily, share the burden of maintenance, and trust that devices are recording actual solar production. As IoT continues to evolve and its adoption continues to grow, the ability to autonomously manage devices and actions taken by devices will be essential. Blockchain and smart contracts are positioned well to integrate those capabilities into IoT.
csong27 / Auditing Text GenerationCode for Auditing Data Provenance in Text-Generation Models (in KDD 2019)
warisgill / TraceFLTraceFL is a novel mechanism for Federated Learning that achieves interpretability by tracking neuron provenance. It identifies clients responsible for global model predictions, achieving 99% accuracy across diverse datasets (e.g., medical imaging) and neural networks (e.g., GPT).
ezozu / CognitiveLedgerFederated learning on-chain via homomorphic encryption for decentralized AI model governance and incentivized data provenance.
provena / ProvenaA provenance system for supporting modelling and simulation workflows
ellennickles / Ml5js Model And Data Provenance ProjectA research project on the origins of ml5.js’ pre-trained models and the data on which they were trained.
humblemat810 / KogwistarGraph-native AI substrate for HyperGraphRAG. Knowledge, Workflow orchestration, conversation memory and wisdom are modeled as hypergraphs. Support strong event sourced provenance and replay.
bprinty / Flask ContinuumModel provenance and versioning via SQLAlchemy-Continuum
timrdf / Prov LodspeakrModels and Views for W3C Provenance Working Group's OWL Ontology
ezozu / CogniChainDecentralized federated learning via homomorphic encryption on-chain; verifiable AI model provenance anchored in immutable directed acyclic graphs.
HDI-Project / Model Provenance JsonA specification for the model provenance file that keeps track of the journey from raw data to deployed model in Machine Learning 2.0 projects.