Concordia: Self-Improving Synthetic Tables for Federated LLMs

Federated learning (FL) enables training large language models (LLMs) without sharing raw data, but adapting LLMs under strict data isolation and non-IID client distributions remains challenging in practice. Synthetic data offers a natural privacy-preserving surrogate for local training, yet existing federated pipelines typically treat synthetic generation as static or loosely coupled with downstream optimization, leading to rapidly diminishing utility under heterogeneous clients. We study federated adaptation of LLMs on tabular tasks where raw records and validation data cannot be shared, and local training must rely entirely on synthetic tables. We propose Concordia, a tri-level optimization framework that aligns synthetic data generation with federated validation utility despite these constraints. At the client level, models are adapted via parameter-efficient LoRA training on synthetic tables. Clients additionally learn lightweight utility scorers from private validation feedback to reweight synthetic samples during local training. At the outer level, each client refines its own synthetic table generator using group-relative policy optimization (GRPO), guided by an ensemble of heterogeneous scorers shared across clients, without aggregating generator parameters or exposing validation data. Experiments on privacy-sensitive tabular benchmarks from finance and healthcare demonstrate that Concordia consistently improves federated performance, cross-client stability, and robustness to distribution shift compared to static and decoupled synthetic-data baselines.

Polarizable atomic multipoles for learning long-range electrostatics

Long-range electrostatics and polarization remain central obstacles to extending machine learning interatomic potentials (MLIPs) to ionic, polar, and interfacial systems. Here, we introduce a semi-local framework for learning electrostatics from energies and forces using polarizable atomic multipoles. Local equivariant descriptors predict environment-dependent latent monopoles, dipoles, and quadrupoles, while residual non-local charge transfer and polarization are captured by non-self-consistent linear response in induced charges and dipoles. Across four diverse benchmarks and four short-range MLIP architectures, the multipole hierarchy and response terms systematically improve potential energy surface accuracy, with the largest gains in systems where long-range effects are essential. More importantly, the learned latent variables recover physically meaningful electrical responses: accurate Born effective charge tensors, emergent polarizabilities, infrared spectra in close agreement with experiments, and semi-quantitative Raman spectra for bulk water and hybrid MAPbI$_3$ perovskite. This systematically improvable, physically transparent framework enables MLIPs trained on standard energy and force labels to predict polarization-sensitive observables.

Efficient Rationale-based Retrieval: On-policy Distillation from Generative Rerankers based on JEPA

Unlike traditional fact-based retrieval, rationale-based retrieval typically necessitates cross-encoding of query-document pairs using large language models, incurring substantial computational costs. To address this limitation, we propose Rabtriever, which independently encodes queries and documents, while providing comparable cross query-document comprehension capabilities to rerankers. We start from training a LLM-based generative reranker, which puts the document prior to the query and prompts the LLM to generate the relevance score by log probabilities. We then employ it as the teacher of an on-policy distillation framework, with Rabtriever as the student to reconstruct the teacher's contextual-aware query embedding. To achieve this effect, Rabtriever is first initialized from the teacher, with parameters frozen. The Joint-Embedding Predictive Architecture (JEPA) paradigm is then adopted, which integrates a lightweight, trainable predictor between LLM layers and heads, projecting the query embedding into a new hidden space, with the document embedding as the latent vector. JEPA then minimizes the distribution difference between this projected embedding and the teacher embedding. To strengthen the sampling efficiency of on-policy distillation, we also add an auxiliary loss on the reverse KL of LLM logits, to reshape the student's logit distribution. Rabtriever optimizes the teacher's quadratic complexity on the document length to linear, verified both theoretically and empirically. Experiments show that Rabtriever outperforms different retriever baselines across diverse rationale-based tasks, including empathetic conversations and robotic manipulations, with minor accuracy degradation from the reranker. Rabtriever also generalizes well on traditional retrieval benchmarks such as MS MARCO and BEIR, with comparable performance to the best retriever baseline.

Breaking Watermarks in the Frequency Domain: A Modulated Diffusion Attack Framework

Digital image watermarking has advanced rapidly for copyright protection of generative AI, yet the comparatively limited progress in watermark attack techniques has broken the attack-defense balance and hindered further advances in the field. In this paper, we propose FMDiffWA, a frequency-domain modulated diffusion framework for watermark attacks. Specifically, we introduce a frequency-domain watermark modulation (FWM) module and incorporate it into the sampling stages both the forward and reverse diffusion processes. This mechanism enables selective modulation of watermark-related frequency components, thereby allowing FMDiffWA to effectively neutralize the invisible watermark signals while preserving the perceptual quality of the attacked watermarked images. To achieve a better trade-off between attack efficacy and visual fidelity, we reformulate the training strategy of conventional diffusion models by augmenting the canonical noise estimation objective with an auxiliary refinement constraint. Comprehensive experiments demonstrate that FMDiffWA achieves superior visual fidelity compared to existing watermark attacks, while exhibiting strong generalization across diverse watermarking schemes.

Byzantine-Robust Distributed SGD: A Unified Analysis and Tight Error Bounds

Byzantine-robust distributed optimization relies on robust aggregation rules to mitigate the influence of malicious Byzantine workers. Despite the proliferation of such rules, a unified convergence analysis framework that accommodates general data heterogeneity is lacking. In this work, we provide a thorough convergence theory of Byzantine-robust distributed stochastic gradient descent (SGD), analyzing variants both with and without local momentum. We establish the convergence rates for nonconvex smooth objectives and those satisfying the Polyak-Lojasiewicz condition under a general data heterogeneity assumption. Our analysis reveals that while stochasticity and data heterogeneity introduce unavoidable error floors, local momentum provably reduces the error component induced by stochasticity. Furthermore, we derive matching lower bounds to demonstrate that the upper bounds obtained in our analysis are tight and characterize the fundamental limits of Byzantine resilience under stochasticity and data heterogeneity. Empirical results support our theoretical findings.

MedCollab: Causal-Driven Multi-Agent Collaboration for Full-Cycle Clinical Diagnosis via IBIS-Structured Argumentation

Large language models (LLMs) have shown promise in healthcare applications, however, their use in clinical practice is still limited by diagnostic hallucinations and insufficiently interpretable reasoning. We present MedCollab, a novel multi-agent framework that emulates the hierarchical consultation workflow of modern hospitals to autonomously navigate the full-cycle diagnostic process. The framework incorporates a dynamic specialist recruitment mechanism that adaptively assembles clinical and examination agents according to patient-specific symptoms and examination results. To ensure the rigor of clinical work, we adopt a structured Issue-Based Information System (IBIS) argumentation protocol that requires agents to provide ``Positions'' backed by traceable evidence from medical knowledge and clinical data. Furthermore, the framework constructs a Hierarchical Disease Causal Chain that transforms flattened diagnostic predictions into a structured model of pathological progression through explicit logical operators. A multi-round Consensus Mechanism iteratively filters low-quality reasoning through logic auditing and weighted voting. Evaluated on real-world clinical datasets, MedCollab significantly outperforms pure LLMs and medical multi-agent systems in Accuracy and RaTEScore, demonstrating a marked reduction in medical hallucinations. These findings indicate that MedCollab provides an extensible, transparent, and clinically compliant approach to medical decision-making.

ERNIE 5.0 Technical Report

In this report, we introduce ERNIE 5.0, a natively autoregressive foundation model desinged for unified multimodal understanding and generation across text, image, video, and audio. All modalities are trained from scratch under a unified next-group-of-tokens prediction objective, based on an ultra-sparse mixture-of-experts (MoE) architecture with modality-agnostic expert routing. To address practical challenges in large-scale deployment under diverse resource constraints, ERNIE 5.0 adopts a novel elastic training paradigm. Within a single pre-training run, the model learns a family of sub-models with varying depths, expert capacities, and routing sparsity, enabling flexible trade-offs among performance, model size, and inference latency in memory- or time-constrained scenarios. Moreover, we systematically address the challenges of scaling reinforcement learning to unified foundation models, thereby guaranteeing efficient and stable post-training under ultra-sparse MoE architectures and diverse multimodal settings. Extensive experiments demonstrate that ERNIE 5.0 achieves strong and balanced performance across multiple modalities. To the best of our knowledge, among publicly disclosed models, ERNIE 5.0 represents the first production-scale realization of a trillion-parameter unified autoregressive model that supports both multimodal understanding and generation. To facilitate further research, we present detailed visualizations of modality-agnostic expert routing in the unified model, alongside comprehensive empirical analysis of elastic training, aiming to offer profound insights to the community.

Stabilizing Decentralized Federated Fine-Tuning via Topology-Aware Alternating LoRA

Decentralized federated learning (DFL), a serverless variant of federated learning, poses unique challenges for parameter-efficient fine-tuning due to the factorized structure of low-rank adaptation (LoRA). Unlike linear parameters, decentralized aggregation of LoRA updates introduces topology-dependent cross terms that can destabilize training under dynamic communication graphs. We propose \texttt{TAD-LoRA}, a Topology-Aware Decentralized Low-Rank Adaptation framework that coordinates the updates and mixing of LoRA factors to control inter-client misalignment. We theoretically prove the convergence of \texttt{TAD-LoRA} under non-convex objectives, explicitly characterizing the trade-off between topology-induced cross-term error and block-coordinate representation bias governed by the switching interval of alternative training. Experiments under various communication conditions validate our analysis, showing that \texttt{TAD-LoRA} achieves robust performance across different communication scenarios, remaining competitive in strongly connected topologies and delivering clear gains under moderately and weakly connected topologies, with particularly strong results on the MNLI dataset.

PID-Guided Partial Alignment for Multimodal Decentralized Federated Learning

Multimodal decentralized federated learning (DFL) is challenging because agents differ in available modalities and model architectures, yet must collaborate over peer-to-peer (P2P) networks without a central coordinator. Standard multimodal pipelines learn a single shared embedding across all modalities. In DFL, such a monolithic representation induces gradient misalignment between uni- and multimodal agents; as a result, it suppresses heterogeneous sharing and cross-modal interaction. We present PARSE, a multimodal DFL framework that operationalizes partial information decomposition (PID) in a server-free setting. Each agent performs feature fission to factorize its latent representation into redundant, unique, and synergistic slices. P2P knowledge sharing among heterogeneous agents is enabled by slice-level partial alignment: only semantically shareable branches are exchanged among agents that possess the corresponding modality. By removing the need for central coordination and gradient surgery, PARSE resolves uni-/multimodal gradient conflicts, thereby overcoming the multimodal DFL dilemma while remaining compatible with standard DFL constraints. Across benchmarks and agent mixes, PARSE yields consistent gains over task-, modality-, and hybrid-sharing DFL baselines. Ablations on fusion operators and split ratios, together with qualitative visualizations, further demonstrate the efficiency and robustness of the proposed design.

CuMA: Aligning LLMs with Sparse Cultural Values via Demographic-Aware Mixture of Adapters

As Large Language Models (LLMs) serve a global audience, alignment must transition from enforcing universal consensus to respecting cultural pluralism. We demonstrate that dense models, when forced to fit conflicting value distributions, suffer from \textbf{Mean Collapse}, converging to a generic average that fails to represent diverse groups. We attribute this to \textbf{Cultural Sparsity}, where gradient interference prevents dense parameters from spanning distinct cultural modes. To resolve this, we propose \textbf{\textsc{CuMA}} (\textbf{Cu}ltural \textbf{M}ixture of \textbf{A}dapters), a framework that frames alignment as a \textbf{conditional capacity separation} problem. By incorporating demographic-aware routing, \textsc{CuMA} internalizes a \textit{Latent Cultural Topology} to explicitly disentangle conflicting gradients into specialized expert subspaces. Extensive evaluations on WorldValuesBench, Community Alignment, and PRISM demonstrate that \textsc{CuMA} achieves state-of-the-art performance, significantly outperforming both dense baselines and semantic-only MoEs. Crucially, our analysis confirms that \textsc{CuMA} effectively mitigates mean collapse, preserving cultural diversity. Our code is available at https://github.com/Throll/CuMA.