Category: News

Parallel Decoder Transformer: Planner-Seeded Latent Coordination for Synchronized Parallel Decodingcs.AI updates on arXiv.org arXiv:2512.10054v2 Announce Type: replace
Abstract: Autoregressive language models can often identify parallel subproblems, but standard decoding exposes only a single left-to-right output interface. External orchestration methods can launch multiple prompts concurrently, yet they provide no model-internal state through which those generations can synchronize, resolve ownership, or wait for missing information. We present the Parallel Decoder Transformer (PDT), a frozen-trunk architecture that augments a decoder with a planner-seeded latent workspace and a synchronized multi-stream output protocol. Before any stream emits tokens, a mandatory prompt-time planner predicts fixed latent plan slots and projects them as snapshot 0 on an embeddings-only Dynamic Notes Bus. During decoding, each stream reads the visible notes window through Speculative Note Conditioning (SNC), emits provisional token blocks and latent summaries, and advances only when agreement logic determines that the current shared state is sufficient for continued parallel generation. Coverage heads track plan-item ownership, while rollback handles incoherent or premature commits. PDT therefore shifts parallel task decomposition from an external prompting strategy to a model-internal coordination mechanism over the output interface of a frozen language model.

 arXiv:2512.10054v2 Announce Type: replace
Abstract: Autoregressive language models can often identify parallel subproblems, but standard decoding exposes only a single left-to-right output interface. External orchestration methods can launch multiple prompts concurrently, yet they provide no model-internal state through which those generations can synchronize, resolve ownership, or wait for missing information. We present the Parallel Decoder Transformer (PDT), a frozen-trunk architecture that augments a decoder with a planner-seeded latent workspace and a synchronized multi-stream output protocol. Before any stream emits tokens, a mandatory prompt-time planner predicts fixed latent plan slots and projects them as snapshot 0 on an embeddings-only Dynamic Notes Bus. During decoding, each stream reads the visible notes window through Speculative Note Conditioning (SNC), emits provisional token blocks and latent summaries, and advances only when agreement logic determines that the current shared state is sufficient for continued parallel generation. Coverage heads track plan-item ownership, while rollback handles incoherent or premature commits. PDT therefore shifts parallel task decomposition from an external prompting strategy to a model-internal coordination mechanism over the output interface of a frozen language model. Read More  

Mitigating Unintended Memorization with LoRA in Federated Learning for LLMscs.AI updates on arXiv.org arXiv:2502.05087v3 Announce Type: replace-cross
Abstract: Federated learning (FL) is a popular paradigm for collaborative training which avoids direct data exposure between clients. However, data privacy issues still remain: FL-trained large language models are capable of memorizing and completing phrases and sentences contained in training data when given their prefixes. Thus, it is possible for adversarial and honest- but-curious clients to recover training data of other participants simply through targeted prompting. In this work, we demonstrate that a popular and simple fine-tuning strategy, low-rank adaptation (LoRA), reduces memorization during FL by a factor of up to 10 without significant performance cost. We study this effect by performing fine-tuning tasks in high-risk domains such as medicine, law, and finance. We observe a reduction in memorization for a wide variety of model families, from 1B to 70B parameters. We find that LoRA can reduce memorization in centralized learning as well, and we compare how the memorization patterns differ. Furthermore, we study the effect of hyperparameters and show that LoRA can be combined with other privacy-preserving techniques such as gradient clipping and Gaussian noise, secure aggregation, and Goldfish loss to further improve record-level privacy while maintaining performance.

 arXiv:2502.05087v3 Announce Type: replace-cross
Abstract: Federated learning (FL) is a popular paradigm for collaborative training which avoids direct data exposure between clients. However, data privacy issues still remain: FL-trained large language models are capable of memorizing and completing phrases and sentences contained in training data when given their prefixes. Thus, it is possible for adversarial and honest- but-curious clients to recover training data of other participants simply through targeted prompting. In this work, we demonstrate that a popular and simple fine-tuning strategy, low-rank adaptation (LoRA), reduces memorization during FL by a factor of up to 10 without significant performance cost. We study this effect by performing fine-tuning tasks in high-risk domains such as medicine, law, and finance. We observe a reduction in memorization for a wide variety of model families, from 1B to 70B parameters. We find that LoRA can reduce memorization in centralized learning as well, and we compare how the memorization patterns differ. Furthermore, we study the effect of hyperparameters and show that LoRA can be combined with other privacy-preserving techniques such as gradient clipping and Gaussian noise, secure aggregation, and Goldfish loss to further improve record-level privacy while maintaining performance. Read More  

Noisy PDE Training Requires Bigger PINNscs.AI updates on arXiv.org arXiv:2507.06967v2 Announce Type: replace-cross
Abstract: Physics-Informed Neural Networks (PINNs) are increasingly used to approximate solutions of partial differential equations (PDEs), particularly in high dimensions. In real-world settings, data are often noisy, making it crucial to understand when a predictor can still achieve low empirical risk. Yet, little is known about the conditions under which a PINN can do so effectively. We analyse PINNs applied to the Hamilton–Jacobi–Bellman (HJB) PDE and establish a lower bound on the network size required for the supervised PINN empirical risk to fall below the variance of noisy supervision labels. Specifically, if a predictor achieves empirical risk $O(eta)$ below $sigma^2$ (the variance of the supervision data), then necessarily $d_Nlog d_Ngtrsim N_s eta^2$, where $N_s$ is the number of samples and $d_N$ the number of trainable parameters. A similar constraint holds in the fully unsupervised PINN setting when boundary labels are noisy. Thus, simply increasing the number of noisy supervision labels does not offer a “free lunch” in reducing empirical risk. We also give empirical studies on the HJB PDE, the Poisson PDE and the the Navier-Stokes PDE set to produce the Taylor-Green solutions. In these experiments we demonstrate that PINNs indeed need to be beyond a threshold model size for them to train to errors below $sigma^2$. These results provide a quantitative foundation for understanding parameter requirements when training PINNs in the presence of noisy data.

 arXiv:2507.06967v2 Announce Type: replace-cross
Abstract: Physics-Informed Neural Networks (PINNs) are increasingly used to approximate solutions of partial differential equations (PDEs), particularly in high dimensions. In real-world settings, data are often noisy, making it crucial to understand when a predictor can still achieve low empirical risk. Yet, little is known about the conditions under which a PINN can do so effectively. We analyse PINNs applied to the Hamilton–Jacobi–Bellman (HJB) PDE and establish a lower bound on the network size required for the supervised PINN empirical risk to fall below the variance of noisy supervision labels. Specifically, if a predictor achieves empirical risk $O(eta)$ below $sigma^2$ (the variance of the supervision data), then necessarily $d_Nlog d_Ngtrsim N_s eta^2$, where $N_s$ is the number of samples and $d_N$ the number of trainable parameters. A similar constraint holds in the fully unsupervised PINN setting when boundary labels are noisy. Thus, simply increasing the number of noisy supervision labels does not offer a “free lunch” in reducing empirical risk. We also give empirical studies on the HJB PDE, the Poisson PDE and the the Navier-Stokes PDE set to produce the Taylor-Green solutions. In these experiments we demonstrate that PINNs indeed need to be beyond a threshold model size for them to train to errors below $sigma^2$. These results provide a quantitative foundation for understanding parameter requirements when training PINNs in the presence of noisy data. Read More  

ByteDance Releases DeerFlow 2.0: An Open-Source SuperAgent Harness that Orchestrates Sub-Agents, Memory, and Sandboxes to do Complex TasksMarkTechPost The era of the ‘Copilot’ is officially getting an upgrade. While the tech world has spent the last two years getting comfortable with AI that suggests code or drafts emails, ByteDance team is moving the goalposts. They released DeerFlow 2.0, a newly open-sourced ‘SuperAgent’ framework that doesn’t just suggest work; it executes it. DeerFlow is
The post ByteDance Releases DeerFlow 2.0: An Open-Source SuperAgent Harness that Orchestrates Sub-Agents, Memory, and Sandboxes to do Complex Tasks appeared first on MarkTechPost.

 The era of the ‘Copilot’ is officially getting an upgrade. While the tech world has spent the last two years getting comfortable with AI that suggests code or drafts emails, ByteDance team is moving the goalposts. They released DeerFlow 2.0, a newly open-sourced ‘SuperAgent’ framework that doesn’t just suggest work; it executes it. DeerFlow is
The post ByteDance Releases DeerFlow 2.0: An Open-Source SuperAgent Harness that Orchestrates Sub-Agents, Memory, and Sandboxes to do Complex Tasks appeared first on MarkTechPost. Read More  

How to Build a Risk-Aware AI Agent with Internal Critic, Self-Consistency Reasoning, and Uncertainty Estimation for Reliable Decision-MakingMarkTechPost In this tutorial, we build an advanced agent system that goes beyond simple response generation by integrating an internal critic and uncertainty estimation framework. We simulate multi-sample inference, evaluate candidate responses across accuracy, coherence, and safety dimensions, and quantify predictive uncertainty using entropy, variance, and consistency measures. We implement risk-sensitive selection strategies to balance confidence
The post How to Build a Risk-Aware AI Agent with Internal Critic, Self-Consistency Reasoning, and Uncertainty Estimation for Reliable Decision-Making appeared first on MarkTechPost.

 In this tutorial, we build an advanced agent system that goes beyond simple response generation by integrating an internal critic and uncertainty estimation framework. We simulate multi-sample inference, evaluate candidate responses across accuracy, coherence, and safety dimensions, and quantify predictive uncertainty using entropy, variance, and consistency measures. We implement risk-sensitive selection strategies to balance confidence
The post How to Build a Risk-Aware AI Agent with Internal Critic, Self-Consistency Reasoning, and Uncertainty Estimation for Reliable Decision-Making appeared first on MarkTechPost. Read More  

Mastercard brings agentic payments to life in Singapore with DBS and UOBAI News Mastercard has completed its first live, authenticated agent-based payment transaction in Singapore, a milestone that advances autonomous AI commerce from proof of concept to everyday use. Announced on March 4, 2026, the transaction was carried out in partnership with DBS and UOB, two of Southeast Asia’s largest banks. In the demonstration, an AI agent booked
The post Mastercard brings agentic payments to life in Singapore with DBS and UOB appeared first on AI News.

 Mastercard has completed its first live, authenticated agent-based payment transaction in Singapore, a milestone that advances autonomous AI commerce from proof of concept to everyday use. Announced on March 4, 2026, the transaction was carried out in partnership with DBS and UOB, two of Southeast Asia’s largest banks. In the demonstration, an AI agent booked
The post Mastercard brings agentic payments to life in Singapore with DBS and UOB appeared first on AI News. Read More  

OpenAI to acquire PromptfooOpenAI News OpenAI is acquiring Promptfoo, an AI security platform that helps enterprises identify and remediate vulnerabilities in AI systems during development.

 OpenAI is acquiring Promptfoo, an AI security platform that helps enterprises identify and remediate vulnerabilities in AI systems during development. Read More  

UK sovereign AI fund to build up domestic computing infrastructureAI News The UK sovereign AI fund intends to secure advantages by providing a domestic alternative to external computing infrastructure. Backed by a £500 million budget from the Department for Science, Innovation and Technology, the unit formally launches on April 16th at 6pm GMT. James Wise, Partner at Balderton Capital, chairs the function to coordinate efforts across
The post UK sovereign AI fund to build up domestic computing infrastructure appeared first on AI News.

 The UK sovereign AI fund intends to secure advantages by providing a domestic alternative to external computing infrastructure. Backed by a £500 million budget from the Department for Science, Innovation and Technology, the unit formally launches on April 16th at 6pm GMT. James Wise, Partner at Balderton Capital, chairs the function to coordinate efforts across
The post UK sovereign AI fund to build up domestic computing infrastructure appeared first on AI News. Read More  

7 Ways People Are Making Money Using AI in 2026KDnuggets Learn how people are turning AI tools into real income by building practical systems, selling outcomes, and creating niche products that businesses are willing to pay for.

 Learn how people are turning AI tools into real income by building practical systems, selling outcomes, and creating niche products that businesses are willing to pay for. Read More