MoLFormers is a suite of molecular foundation models that accelerate the development of chemicals, materials, and biologics. It includes a number of state-of-the-art pre-trained and fine-tuned models for molecular data analysis, property prediction, nearest neighbor retrieval, and discovery (such as design of catalysts, stability agents, antibodies, protein inhibitors) tasks, thus enabling molecular science at higher accuracy and success rate, lower cost, and faster speed. The suite contains high-quality families (encoder-only, encoder-decoder, decoder-only) of large molecular models that combine best-of-breed architectures (e.g., rotary embedding, multi-modal embedding, prompt tuning, linear attention, geometry encoder, etc.) and provide world-class predictive and generative AI capabilities. One such instantiation of MoLFormer is MoLFormer-XL, a chemical SMILES encoder trained on 1.1 B chemicals using rotary positional encodings and linear attention. MoLFormer-XL, finetuned on downstr
With the accelerated advances in Generative AI, Gartner predicted that by 2030, synthetic data will overshadow real data in training AI models. The primary motivation for using synthetic data in the AI life cycle comes from its promise in alleviating risk, bias, harm, and privacy concerns inherent in the real data. How can we guarantee that such an approach delivers on its promises? How can we determine that a synthetic dataset indeed conforms to fairness, privacy, robustness and fidelity requirements, and is still useful to downstream tasks?
While Foundation Models (FM) have greatly transformed AI solutions for language and vision, they often fall short in addressing time-series data, which is widely used in various industries. At IBM Research, our dedicated team focuses exclusively on advancing Time Series Foundation Models and has made significant contributions with several influential papers presented at top AI conferences. Our team has been pioneers in this space where we defined the first inaugural architecture for several popular Time-series FM backbones, including the first transformer for multi-variate time-series representation learning (TST, KDD 21), the first patched time-series transformer (PatchTST, ICLR 23) and the first patched MLP-Mixer for time series (PatchTSMixer, KDD 23). Our latest Models (PatchTST and PatchTSMixer) are the leading SOTAs in this space with a significant reduction (2-3X) in compute and memory requirements. We have released our SOTA models through various open-source channels attracting
Traditional data integration techniques often require complex coding and a deep understanding of data architectures, which can be daunting for non-specialists. In the evolving landscape of AI, there's a growing need for tools that democratize data access and analysis. We present FlowPilot, a novel system that departs from the current one-shot text-to-SQL paradigms that often fail to answer complex queries.
The emergence of foundation models has significantly lowered the barriers to applying AI to everyday problems, transforming the way organizations consume, customize and build AI-enabled applications. We are also seeing the emergence of a new persona, the AI Builder, who is in need of dedicated tooling to harness the power of LLMs while mitigating its associated risks.
MoLFormers is a suite of molecular foundation models that accelerate the development of chemicals, materials, and biologics. It includes a number of state-of-the-art pre-trained and fine-tuned models for molecular data analysis, property prediction, nearest neighbor retrieval, and discovery (such as design of catalysts, stability agents, antibodies, protein inhibitors) tasks, thus enabling molecular science at higher accuracy and success rate, lower cost, and faster speed. The suite contains high-quality families (encoder-only, encoder-decoder, decoder-only) of large molecular models that combine best-of-breed architectures (e.g., rotary embedding, multi-modal embedding, prompt tuning, linear attention, geometry encoder, etc.) and provide world-class predictive and generative AI capabilities. One such instantiation of MoLFormer is MoLFormer-XL, a chemical SMILES encoder trained on 1.1 B chemicals using rotary positional encodings and linear attention. MoLFormer-XL, finetuned on downstr
The fast-increasing complexity of modern IT in multi cloud environments is bringing unprecedented management challenges to Site Reliability Engineers (SREs) to meet Service Level Objectives (SLOs) and keep systems up and running effectively. To put in perspective, an availability SLO of 99.99% allows for 4.3 minutes of downtime per month, hardly something that can be attained by simply reacting to incidents. In this demo, we introduce our approach to address this challenge by transforming ITOps from being reactive to becoming proactive by leveraging large language models and advanced AI capabilities. The main goal of our work is to automate as much as possible the implementation of resolutions for upcoming IT issues before they turn into outages. Our demo consists of four steps:
(1) Issue Detection, where we have developed an unsupervised methodology for detecting issues via ensemble of various anomaly detectors. We compare our methods with the state-of-the-art techniques implemented i
This demo shows RADAR, a language model trained to detect AI-written text generated from large language models (LLMs). The live demo consists of four RADAR detectors trained on different LLMs. Users can paste in a paragraph of text and receive likelihood predictions of the AI-written text in real time. Examples of AI versus human written text are also provided in the demo. The demo is based on the NeurIPS 2023 paper "RADAR: Robust AI-Text Detection via Adversarial Learning".
With the wide range of new applications using machine learning models, and the recent disruptive power of foundation models, the robustness and trustworthiness of artificial intelligence has become imperative. We will present the Adversarial Robustness Toolbox (ART), an open-source project which is part of The Linux Foundation AI & Data, that provides tools to assess and improve the security of machine learning models. Recently, ART supports for models hosted by Hugging Face, and as such, we will present a demonstration of how to use this new functionality.
With the wide range of new applications using machine learning models, and the recent disruptive power of foundation models, the robustness and trustworthiness of artificial intelligence has become imperative. We will present the Adversarial Robustness Toolbox (ART), an open-source project which is part of The Linux Foundation AI & Data, that provides tools to assess and improve the security of machine learning models. Recently, ART supports for models hosted by Hugging Face, and as such, we will present a demonstration of how to use this new functionality.
With the accelerated advances in Generative AI, Gartner predicted that by 2030, synthetic data will overshadow real data in training AI models. The primary motivation for using synthetic data in the AI life cycle comes from its promise in alleviating risk, bias, harm, and privacy concerns inherent in the real data. How can we guarantee that such an approach delivers on its promises? How can we determine that a synthetic dataset indeed conforms to fairness, privacy, robustness and fidelity requirements, and is still useful to downstream tasks?
MoLFormers is a suite of molecular foundation models that accelerate the development of chemicals, materials, and biologics. It includes a number of state-of-the-art pre-trained and fine-tuned models for molecular data analysis, property prediction, nearest neighbor retrieval, and discovery (such as design of catalysts, stability agents, antibodies, protein inhibitors) tasks, thus enabling molecular science at higher accuracy and success rate, lower cost, and faster speed. The suite contains high-quality families (encoder-only, encoder-decoder, decoder-only) of large molecular models that combine best-of-breed architectures (e.g., rotary embedding, multi-modal embedding, prompt tuning, linear attention, geometry encoder, etc.) and provide world-class predictive and generative AI capabilities. One such instantiation of MoLFormer is MoLFormer-XL, a chemical SMILES encoder trained on 1.1 B chemicals using rotary positional encodings and linear attention. MoLFormer-XL, finetuned on downstr
The emergence of foundation models has significantly lowered the barriers to applying AI to everyday problems, transforming the way organizations consume, customize and build AI-enabled applications. We are also seeing the emergence of a new persona, the AI Builder, who is in need of dedicated tooling to harness the power of LLMs while mitigating its associated risks.
Traditional data integration techniques often require complex coding and a deep understanding of data architectures, which can be daunting for non-specialists. In the evolving landscape of AI, there's a growing need for tools that democratize data access and analysis. We present FlowPilot, a novel system that departs from the current one-shot text-to-SQL paradigms that often fail to answer complex queries.
While Foundation Models (FM) have greatly transformed AI solutions for language and vision, they often fall short in addressing time-series data, which is widely used in various industries. At IBM Research, our dedicated team focuses exclusively on advancing Time Series Foundation Models and has made significant contributions with several influential papers presented at top AI conferences. Our team has been pioneers in this space where we defined the first inaugural architecture for several popular Time-series FM backbones, including the first transformer for multi-variate time-series representation learning (TST, KDD 21), the first patched time-series transformer (PatchTST, ICLR 23) and the first patched MLP-Mixer for time series (PatchTSMixer, KDD 23). Our latest Models (PatchTST and PatchTSMixer) are the leading SOTAs in this space with a significant reduction (2-3X) in compute and memory requirements. We have released our SOTA models through various open-source channels attracting
This demo shows RADAR, a language model trained to detect AI-written text generated from large language models (LLMs). The live demo consists of four RADAR detectors trained on different LLMs. Users can paste in a paragraph of text and receive likelihood predictions of the AI-written text in real time. Examples of AI versus human written text are also provided in the demo. The demo is based on the NeurIPS 2023 paper "RADAR: Robust AI-Text Detection via Adversarial Learning".
The fast-increasing complexity of modern IT in multi cloud environments is bringing unprecedented management challenges to Site Reliability Engineers (SREs) to meet Service Level Objectives (SLOs) and keep systems up and running effectively. To put in perspective, an availability SLO of 99.99% allows for 4.3 minutes of downtime per month, hardly something that can be attained by simply reacting to incidents. In this demo, we introduce our approach to address this challenge by transforming ITOps from being reactive to becoming proactive by leveraging large language models and advanced AI capabilities. The main goal of our work is to automate as much as possible the implementation of resolutions for upcoming IT issues before they turn into outages. Our demo consists of four steps:
(1) Issue Detection, where we have developed an unsupervised methodology for detecting issues via ensemble of various anomaly detectors. We compare our methods with the state-of-the-art techniques implemented i
With the wide range of new applications using machine learning models, and the recent disruptive power of foundation models, the robustness and trustworthiness of artificial intelligence has become imperative. We will present the Adversarial Robustness Toolbox (ART), an open-source project which is part of The Linux Foundation AI & Data, that provides tools to assess and improve the security of machine learning models. Recently, ART supports for models hosted by Hugging Face, and as such, we will present a demonstration of how to use this new functionality.
Traditional data integration techniques often require complex coding and a deep understanding of data architectures, which can be daunting for non-specialists. In the evolving landscape of AI, there's a growing need for tools that democratize data access and analysis. We present FlowPilot, a novel system that departs from the current one-shot text-to-SQL paradigms that often fail to answer complex queries.
MoLFormers is a suite of molecular foundation models that accelerate the development of chemicals, materials, and biologics. It includes a number of state-of-the-art pre-trained and fine-tuned models for molecular data analysis, property prediction, nearest neighbor retrieval, and discovery (such as design of catalysts, stability agents, antibodies, protein inhibitors) tasks, thus enabling molecular science at higher accuracy and success rate, lower cost, and faster speed. The suite contains high-quality families (encoder-only, encoder-decoder, decoder-only) of large molecular models that combine best-of-breed architectures (e.g., rotary embedding, multi-modal embedding, prompt tuning, linear attention, geometry encoder, etc.) and provide world-class predictive and generative AI capabilities. One such instantiation of MoLFormer is MoLFormer-XL, a chemical SMILES encoder trained on 1.1 B chemicals using rotary positional encodings and linear attention. MoLFormer-XL, finetuned on downstr
While Foundation Models (FM) have greatly transformed AI solutions for language and vision, they often fall short in addressing time-series data, which is widely used in various industries. At IBM Research, our dedicated team focuses exclusively on advancing Time Series Foundation Models and has made significant contributions with several influential papers presented at top AI conferences. Our team has been pioneers in this space where we defined the first inaugural architecture for several popular Time-series FM backbones, including the first transformer for multi-variate time-series representation learning (TST, KDD 21), the first patched time-series transformer (PatchTST, ICLR 23) and the first patched MLP-Mixer for time series (PatchTSMixer, KDD 23). Our latest Models (PatchTST and PatchTSMixer) are the leading SOTAs in this space with a significant reduction (2-3X) in compute and memory requirements. We have released our SOTA models through various open-source channels attracting
This demo shows RADAR, a language model trained to detect AI-written text generated from large language models (LLMs). The live demo consists of four RADAR detectors trained on different LLMs. Users can paste in a paragraph of text and receive likelihood predictions of the AI-written text in real time. Examples of AI versus human written text are also provided in the demo. The demo is based on the NeurIPS 2023 paper "RADAR: Robust AI-Text Detection via Adversarial Learning".
With the accelerated advances in Generative AI, Gartner predicted that by 2030, synthetic data will overshadow real data in training AI models. The primary motivation for using synthetic data in the AI life cycle comes from its promise in alleviating risk, bias, harm, and privacy concerns inherent in the real data. How can we guarantee that such an approach delivers on its promises? How can we determine that a synthetic dataset indeed conforms to fairness, privacy, robustness and fidelity requirements, and is still useful to downstream tasks?
The emergence of foundation models has significantly lowered the barriers to applying AI to everyday problems, transforming the way organizations consume, customize and build AI-enabled applications. We are also seeing the emergence of a new persona, the AI Builder, who is in need of dedicated tooling to harness the power of LLMs while mitigating its associated risks.
The fast-increasing complexity of modern IT in multi cloud environments is bringing unprecedented management challenges to Site Reliability Engineers (SREs) to meet Service Level Objectives (SLOs) and keep systems up and running effectively. To put in perspective, an availability SLO of 99.99% allows for 4.3 minutes of downtime per month, hardly something that can be attained by simply reacting to incidents. In this demo, we introduce our approach to address this challenge by transforming ITOps from being reactive to becoming proactive by leveraging large language models and advanced AI capabilities. The main goal of our work is to automate as much as possible the implementation of resolutions for upcoming IT issues before they turn into outages. Our demo consists of four steps:
(1) Issue Detection, where we have developed an unsupervised methodology for detecting issues via ensemble of various anomaly detectors. We compare our methods with the state-of-the-art techniques implemented i
While Foundation Models (FM) have greatly transformed AI solutions for language and vision, they often fall short in addressing time-series data, which is widely used in various industries. At IBM Research, our dedicated team focuses exclusively on advancing Time Series Foundation Models and has made significant contributions with several influential papers presented at top AI conferences. Our team has been pioneers in this space where we defined the first inaugural architecture for several popular Time-series FM backbones, including the first transformer for multi-variate time-series representation learning (TST, KDD 21), the first patched time-series transformer (PatchTST, ICLR 23) and the first patched MLP-Mixer for time series (PatchTSMixer, KDD 23). Our latest Models (PatchTST and PatchTSMixer) are the leading SOTAs in this space with a significant reduction (2-3X) in compute and memory requirements. We have released our SOTA models through various open-source channels attracting
The emergence of foundation models has significantly lowered the barriers to applying AI to everyday problems, transforming the way organizations consume, customize and build AI-enabled applications. We are also seeing the emergence of a new persona, the AI Builder, who is in need of dedicated tooling to harness the power of LLMs while mitigating its associated risks.
The fast-increasing complexity of modern IT in multi cloud environments is bringing unprecedented management challenges to Site Reliability Engineers (SREs) to meet Service Level Objectives (SLOs) and keep systems up and running effectively. To put in perspective, an availability SLO of 99.99% allows for 4.3 minutes of downtime per month, hardly something that can be attained by simply reacting to incidents. In this demo, we introduce our approach to address this challenge by transforming ITOps from being reactive to becoming proactive by leveraging large language models and advanced AI capabilities. The main goal of our work is to automate as much as possible the implementation of resolutions for upcoming IT issues before they turn into outages. Our demo consists of four steps:
(1) Issue Detection, where we have developed an unsupervised methodology for detecting issues via ensemble of various anomaly detectors. We compare our methods with the state-of-the-art techniques implemented i
Traditional data integration techniques often require complex coding and a deep understanding of data architectures, which can be daunting for non-specialists. In the evolving landscape of AI, there's a growing need for tools that democratize data access and analysis. We present FlowPilot, a novel system that departs from the current one-shot text-to-SQL paradigms that often fail to answer complex queries.
This demo shows RADAR, a language model trained to detect AI-written text generated from large language models (LLMs). The live demo consists of four RADAR detectors trained on different LLMs. Users can paste in a paragraph of text and receive likelihood predictions of the AI-written text in real time. Examples of AI versus human written text are also provided in the demo. The demo is based on the NeurIPS 2023 paper "RADAR: Robust AI-Text Detection via Adversarial Learning".