Topic
algorithm
Multi-Agent RL System MAMO Automates Weight Selection for Constrained Optimization Problems
MAMO decouples task execution from objective design using multi-agent RL to automatically select reward weights for constrained optimization, improving adaptability in dynamic environments.
New AI Framework PEGE Boosts HIV Detection by 15.4% in Networked Testing
A new AI framework called Policy-Embedded Graph Expansion (PEGE), combined with Dynamics-Driven Branching (DDB), improves HIV detection by 15.4% in networked testing. Developed with WHO and University of Witwatersrand, the approach supports UN Sustainable Development Goal 3.3 by making testing more efficient on incrementally revealed disease networks.
AIGB-Pearl: New AI Method Combines Generative Planning and Policy Optimization for Auto-bidding
Researchers propose AIGB-Pearl, a method that integrates generative planning and policy optimization to overcome performance bottlenecks in AI-generated auto-bidding. The approach uses a trajectory evaluator and a KL-Lipschitz-constrained score-maximization scheme for safe exploration beyond offline datasets, demonstrating state-of-the-art results in simulated and real-world advertising systems.
FastMix: Gradient-Based Data Mixture Optimization Reduces Search Cost in AI Training
FastMix is a novel framework that automates data mixture discovery by training only a single proxy model and jointly optimizing mixture coefficients and model parameters via gradient descent. It reformulates mixture selection as a bilevel optimization problem, enabling efficient, scalable optimization that outperforms baselines.
Fast-dLLM++ Boosts Diffusion LLM Inference Up to 37% With Fréchet Profile Decoding
Researchers propose Fast-dLLM++, a training-free extension to Fast-dLLM that uses Fréchet profile decoding to select parallel token commit sets from the full confidence profile. Experiments on LLaDA-8B show up to 37% higher throughput at comparable accuracy on benchmarks including GSM8K, MATH, HumanEval, and MBPP.
RidgeCut: Reinforcement Learning Framework Optimizes Logistics Network Partitioning with Rings and Wedges
Researchers have developed RidgeCut, a reinforcement learning framework that leverages ring-and-wedge topology to improve graph partitioning for transportation networks. The method consistently outperforms existing approaches in normalized cut metrics and generalizes across graph sizes, offering potential applications in logistics and supply chain network design.
Study Finds Textual Reviews Add Limited Value to Matrix Factorization Recommendations
Researchers systematically evaluated the impact of incorporating textual reviews into matrix factorization for recommendations. They found that adaptive fusion mechanisms improve flexibility, but collaborative signals still dominate performance.
Adaptive kNN Graph Model Decouples Inference Latency from Complexity, Achieving Real-Time Classification
Researchers present an adaptive k-nearest neighbors graph model that decouples inference latency from computational complexity by integrating a Hierarchical Navigable Small World (HNSW) graph with a pre-computed voting mechanism. Benchmarking against eight baselines across six datasets shows real-time performance without compromising classification accuracy.
MeEvo: Metacognitive Evolution Combined with Natural Evolution for Automatic Heuristic Design
A new AI framework called MeEvo cyclically couples natural evolution and metacognitive evolution for automatic heuristic design. It addresses limitations of existing LLM-based approaches by combining population exploration with reflective refinement. Experiments show stronger and more stable performance on complex optimization tasks.
Surprise-Guided MergeSort Reduces Human Ranking Costs by Using AI to Prioritise Comparisons
Researchers propose Surprise-Guided MergeSort (SGS), a framework that combines a Vision-Language Model with MergeSort to schedule pairwise comparisons for subjective ranking tasks. SGS routes only ambiguous comparisons to humans, achieving Kendall's τ×100 improvements of +6 to +12 over Active Elo under the same budget.
Proximal Policy Optimization Achieves Faster Convergence in Discrete Sampling Research
A new paper on arXiv explores policy gradient algorithms for training stochastic policies under the Generative Flow Network (GFlowNet) framework. The authors derive equivalents of standard policy gradient algorithms and, for the first time, successfully apply proximal policy optimization (PPO) to GFlowNets, demonstrating improved convergence speed and data efficiency on benchmarks including synthetic energies and molecular graph generation.
Tensor-Coord: Algebraic Decomposition Enables Conflict-Free Multi-Agent LLM Planning
A new research paper introduces Tensor-Coord, a multilinear algebra framework that represents joint plans of multiple LLM agents as a third-order tensor. By decomposing the tensor, it identifies coordination conflicts and enables iterative replanning, achieving 100% conflict-free plans for 2-agent tasks and 80% for 3-agent tasks in simulated delivery scenarios.
First Wasserstein-2 Convergence Proof for Decentralized Diffusion Models with ODE Samplers
A team of researchers has proven the first convergence guarantee in Wasserstein-2 distance for ODE-based samplers in decentralized diffusion models. The work addresses the missing theoretical foundation for decentralized generative architectures that replace a single global velocity field with multiple local experts and a routing mechanism. The result shows distribution converges at rate O(N^{-1/2}+ε), paving the way for privacy-scalable AI deployments.
Technology Instagram expands algorithm personalization, but not for accounts you follow
Instagram has expanded its algorithm personalization feature to the main feed, allowing users to control which topics they see more or less of. However, the feature does not support requests to see more posts from accounts the user follows, a limitation that frustrates creators and businesses. Instagram chief Adam Mosseri explained that the change is powered by large language models and aims to give users more agency, but acknowledged that the decline of the 'following' feed was a consequence of shifting user behavior.
Technology Jeff Bezos Funds Flourish's $2.5 Billion Quest for a Synthetic Brain That Runs on 50 Watts
Jeff Bezos has invested $50 million (later nearly doubled) into Flourish, a neuro AI startup founded by former Amazon executive Rob Williams and neuroscientist Thomas Reardon. The company aims to build a synthetic intelligence system called Cortex AI that matches the human brain's learning efficiency and power budget of 50 watts or less, addressing the energy inefficiency of large language models. Flourish has raised $500 million at a $2.5 billion valuation from investors including Lux Capital, Google Ventures, and Catalio.