arXiv:2605.28069v1 Announce Type: new
Abstract: Adaptive context compression is vital for scaling Large Language Models (LLMs) to complex, multi-turn agent tasks. However, rule-based compression methods may discard task-critical nuances, while Reinforcement Learning (RL) approaches usually struggle to balance information retention and token efficiency under the sparse rewards inherent to long-horizon workflows. To bridge this gap, we propose ZipRL, a novel adaptive compression framework tailored for Reinforcement Learning from Verifiable Rewards (RLVR). ZipRL features a multi-granularity compression mechanism for active, non-uniform information reduction, coupled with Hindsight Response Replay (HRR), a technique designed to densify training signals during RLVR optimization. Theoretically, we prove ZipRL's superior task-relevant utility over uniform methods. Concretely, ZipRL utilizes coarse-to-fine prompts for macro-compression and incorporates HRR into GRPO via generalized advantage reshaping. Multiple models of varying versions and parameter scales validate the effectiveness of our approach. Benchmarks on five agent tasks show ZipRL outperforms state-of-the-art approaches by 27.9% and 34.7% across Qwen3-4B and Qwen3-8B models, while maintaining exceptional token efficiency and robustness under extreme 256-turn extrapolation stress tests.
Science Journals
arXiv:2605.27607v1 Announce Type: new
Abstract: Quantum squeezed states of light can enhance measurement sensitivity beyond classical limits and enable quantum information processing, but scalable low-loss sources remain challenging. We demonstrate continuous-wave quantum squeezing on a chip, achieving 18 dB of squeezing and 20 dB of anti-squeezing at 1570 nm in a 1.6-cm traveling-wave adaptively poled thin-film lithium niobate waveguide. A distributed model independently determines facet losses, phase noise, and nonlinear interaction strength without prior assumptions, enabling rigorous inference of on-chip performance. We estimate a 95% confidence interval of [-18.96, -17.25] dB squeezing and [19.96, 21.35] dB anti-squeezing. These values represent the highest squeezing reported for any integrated photonic platform and the first assumption-free statistical validation of integrated squeezing performance. Our results establish thin-film lithium niobate as a high-performance, scalable platform for continuous-variable quantum sensing, communications, and photonic computing.
arXiv:2605.27609v1 Announce Type: new
Abstract: We investigate how proximity to a metal nanostructure, particularly to a flat mirror or a cavity confined between two mirrors, affects the vibronic structure of Raman scattering signals. We find that such proximity, particularly for the strong-coupling situation encountered in cavity environments, plays multiple roles in shaping Raman signals beyond the now-familiar signal enhancement known as surface-enhanced Raman scattering (SERS). First, in analogy to the electromagnetic SERS mechanism, near or between mirrors, the local field experienced by a molecule differ from that in vacuum. In particular, between mirrors, the cavity enhances the effective excited state population by trapping the EM field inside it. Second, the nearby metal surface provides a relaxation channel and a lineshape broadening mechanism, and inside a cavity this lineshape is inherited by the cavity polaritons. This relaxation results in a loss of yield but the associated broadening also leads to significant absorption over a larger frequency range. Third, near metallic interfaces interference between incident and reflected light can lead to a richly structured Raman spectrum. For instance, we find that the Rabi contraction (that results from depopulating the ground state) can interfere with Raman signals (and the effect appears to be the same order as Raman itself). These cavity effects are calculated by a full-scale FDTD simulation and highlight the convoluted but fascinating roles of photonic materials on optical signals.
arXiv:2605.27620v1 Announce Type: new
Abstract: A real-time multicore system requires delay bounds on access to shared resources. These resources include the kernel, which has potentially many non-preemptible critical sections guarded by one or more different synchronization primitives. While primitives such as FIFO locks bound the waiting time to enter a critical section, they do not distinguish the importance of individual tasks competing for shared resource access. To address this, we consider a priority-aware spinlock, which reduces the average delay of more important tasks while maintaining a worst-case bound on lock waiting time. We propose a Batched Priority Lock (BPL) that first groups waiting tasks based on the order of their lock requests, and then determines the next lock holder according to priority within the waiting group. We compare BPL to alternative lock approaches, showing that the average waiting time is reduced for higher priority tasks, in simulations up to 64 cores, and for a working implementation on an 8-core machine with a real RTOS. BPL is a compromise between strict priority and FIFO ordering. While strict priorities may lead to starvation and, hence, unbounded lock acquisition delays, BPL has the same waiting bound as FIFO, but with benefits to higher priority tasks. Although its complexity is greater than that of a simple spinlock, its common case execution overhead is shown to be inexpensive in a working system. We believe this is an acceptable cost in systems that require predictability.
arXiv:2605.27626v1 Announce Type: new
Abstract: Geometrically-resolved Direct Numerical Simulations of vertical-axis wind turbines are presented. Simulations were performed using the spectral/hp element method framework Nektar++ with a moving reference frame formulation. Three turbine geometries are considered with one, two and three blades. The full dynamic stall process is resolved, including the formation of large-scale vortices, the separation from the blade, and interaction with the near wake. Increasing blade number introduces blade-vortex interactions that interact with the dynamic stall process. For the three-bladed configuration, these interactions coincide with the early phase of dynamic stall during which the laminar separation bubble develops, and the resulting dynamic stall vortex is reduced in size. Further, the DNS outcomes identified a novel complementary mechanism, where direct vortex impingement causes the premature breakup of the dynamic stall vortices. With the dynamic stall vortex a defining feature of the VAWT near-wake, its accelerated breakdown removes the flow structures that distinguish the VAWT wake from that of a bluff body. Hence the near-wake transitions more rapidly towards bluff-body dynamics, with shear-layer-associated recovery. Self-similarity analysis is extended into the near-wake to quantify this transition, capturing the downstream rate at which the wake loses its dependence on blade-generated coherent structures and collapses onto a self-similar solution. Blade number is shown to be more influential than tip-speed ratio in setting the rate of this transition. The results have implications for closely-spaced turbine arrays and coupled-pair configurations, where the inflow experienced by a downstream rotor is shown to be blade-number-dependent.
arXiv:2605.28089v1 Announce Type: new
Abstract: BuddyBench introduces a privacy-constrained multi-task benchmark for pediatric social-communication personalization. Unlike existing neurodevelopmental repositories that primarily emphasize imaging, genetics, or cross-sectional clinical phenotyping, BuddyBench links drill-level learning trajectories, standardized clinical assessments, BuddyPlan self-report, and randomized-treatment endpoints within a unified benchmark schema. BuddyBench combines two cohorts: ND-03 is an observational cohort with dense drill coverage for Tasks1-2 (n = 189), and ND-02 is a randomized controlled trial cohort for Tasks3-4 (n = 86 ITT). Together, they support knowledge tracing, next-drill recommendation, clinical prediction, and causal inference, linking behavioral personalization to clinical evaluation. We additionally introduce BuddyBench-Sim, a synthetic companion dataset for reproducible evaluation. Baselines show signal across tasks while keeping pediatric clinical records protected.
arXiv:2605.28101v1 Announce Type: new
Abstract: Predicting spatially varying Room Impulse Response (RIR) from sparse observations is a critical but highly challenging inverse problem for immersive spatial audio rendering. In this work, we present EIGENET, a geometry-informed multi-modal framework for few-shot novel view RIR prediction. At its core is a Cross-view Alternate-attention Transformer that iteratively refines local intra-view acoustic structures and global cross-view spatial relationships. We empirically demonstrate that this architecture is capable of making full use of the multi-view multi-modal context while performing spatial-temporal reasoning for RIR prediction. Inspired by acoustic ray tracing, we design a geometry-informed modulation block to formulate the connection between geometric features and RIR power spectrum. In the mean time, an auxiliary loss is introduced to transform the single-target waveform prediction into a multi-task learning framework. Through ablation studies, we demonstrate that this design yields consistent performance gains regardless of the underlying backbone, thereby confirming its foundational utility and architecture-agnostic generalizability for RIR prediction task. Evaluated on both simulated and real-world benchmarks, EIGENET achieves both state-of-the-art performance in few-shot novel view RIR prediction and sim-to-real generalization. Codes and checkpoints are available on https://github.com/FEAfeatherTHER/EigeNet.
arXiv:2605.27459v1 Announce Type: cross
Abstract: Simulation of post-prandial pharmacokinetics, such as muscle protein synthesis (MPS) through mTORC1 and insulin-induced glucose uptake, is often challenging due to the computational intensity of the multi-compartmental approach. In this study, I introduce an in silico metabolic simulator that uses bi-compartmental Bateman kinetic processes, gamma-variate distributions, and finite state machine reasoning to solve temporal differential equations instantaneously, generating metabolic curves and predictions depending on input meals. The novel underlying algorithm was custom-built entirely independent of third-party libraries or external services. This original computational engine, bridging the gap between academia and the digital health sector, is integrated within a web dashboard and provided as a service via REST APIs. The average response time is approximately 135 ms with a maximum below 750 ms. The multi-dimensional model was calibrated using a Landmark Validation approach across diverse dietary conditions (Whey Protein, mixed meal, OGTT) and optimized via Grid Search. Ultimately, the system achieved a global physiologically optimal Mean Absolute Percentage Error (MAPE) of $\sim18\%$ while maintaining an algorithmic complexity of $O(n \log n)$.
arXiv:2605.27668v1 Announce Type: new
Abstract: Probabilistic forecasting estimates the likelihood of uncertain future events. To improve LLM forecasting, existing methods typically learn from binary outcomes to output verbalized forecasts. However, while aggregated human forecasts contain rich information in both the crowd probability estimate and the degree of agreement among forecasters, how to utilize these signals remains underexplored. To address this, we propose the Beta-Bernoulli Calibrator (BBC), which converts an initial point estimate forecast from any model into a distribution over event likelihood, using supervision from both binary outcomes and human forecasts. BBC models event likelihood $p \sim \text{Beta}(\alpha, \beta)$ and outcome $y \sim \text{Bernoulli}(p)$, with the mean as the calibrated point forecast and the variance as the epistemic uncertainty. Our results show that BBC generally provides better calibrated and more accurate forecasts than both traditional post-hoc calibration methods and models fine-tuned specifically for forecasting, while remaining lightweight and having good generalization. We also show that the epistemic uncertainty captured by BBC is a more reliable predictor of forecasting error than verbalized confidence.
arXiv:2605.28122v1 Announce Type: new
Abstract: A coding agent executes a benign task as a sequence of shell, file, and network actions, any of which can quietly exceed the authorized scope while the task still completes. We call this overeager behavior: the prompt is not adversarial and the run succeeds, yet an out-of-scope step can leak credentials or delete files. Existing benchmarks miss it: task-completion suites credit any finished run, jailbreak suites probe adversarial prompts, and the one prior overeager benchmark applies a single fixed prompt set to every agent-model pair, leaving its easiest and most resistant pairs under-measured. We present SNARE (Synthesizing Non-adversarial scenarios for Adaptive Reward-guided Elicitation), a pipeline that composes benign scenarios from reusable scope and trap fragments, scores each run with a judge-free oracle flagging trap-pattern matches and unsolicited file additions or deletions, and uses Thompson sampling to steer each pair's run budget toward the scenarios that most often trigger it. Instantiating it over 24 overeager archetypes yields OverEager, which we run across a 4x5 matrix of four coding agents and five base models. Across 10,000 benign runs, 19.51% trigger overeager behavior, with per-pair rates spanning 11.9x. This variation is driven by the agent framework, not the model: the framework accounts for 56% of it against the model's 21%, so any single-framework or single-model evaluation undercounts the matrix by about a fifth.
arXiv:2605.28157v1 Announce Type: new
Abstract: This study developed a computer-aided diagnosis (CAD) system for detecting caries and molar-incisor hypomineralization (MIH) in intraoral photographs. These lesions share similar appearances, making clinical differentiation challenging, especially given their small size and variability in imaging conditions.
arXiv:2605.28161v1 Announce Type: new
Abstract: Clinical diagnosis of meniscus injuries requires radiologists to integrate volumetric MRI evidence with patient context (e.g., sex, age, BMI) and to produce structured diagnostic reports. Existing knee MRI benchmarks are typically unimodal and rely on coarse labels, limiting their ability to evaluate holistic clinical reasoning. We introduce MeniOmni, a structured multimodal benchmark for meniscus injury assessment, consisting of 746 multi-center MRI studies with tri-planar volumetric inputs, Clinical Priors, and expert-annotated clinical text. MeniOmni supports two tasks: (1) fine-grained Stoller severity grading and (2) diagnostic report generation. We further propose risk-aware ordinal evaluation and a semantic consistency metric (Meni-Score) to better reflect clinical relevance. Baseline experiments show that incorporating Clinical Priors improves grading performance and reduces severe errors, highlighting the value of multimodal context for safer assessment. Code and data are available at https://github.com/ShuruiXu/MeniOmni.
arXiv:2605.27720v1 Announce Type: new
Abstract: Reinforcement learning and data-driven autonomous controllers are commonly evaluated using cumulative reward and empirical success frequency under finite simulation trajectories. However, such empirical metrics do not necessarily provide sufficient statistical evidence regarding deployment readiness under uncertainty. This work develops a Bayesian approval framework for learned autonomous landing controllers under finite rollout evidence. A probabilistic landing capability formulation is introduced based on touchdown safety satisfaction under uncertain operating conditions, while Bayesian posterior inference is used to quantify uncertainty regarding the true deployment capability of learned policies. Posterior approval probability and posterior deployment risk are further introduced for deployment-oriented evaluation, together with a sequential validation framework supporting approve/reject/continue decisions during progressive rollout testing. Simulation experiments using PPO and SAC controllers demonstrate that empirical success and reward optimization may produce overconfident deployment interpretation under limited validation evidence, whereas posterior approval inference provides a more uncertainty-calibrated assessment of deployment readiness. The proposed framework provides a practical statistical connection between conventional reinforcement-learning evaluation and deployment-oriented validation under uncertainty and may be generalized to broader classes of learned autonomous systems.
arXiv:2605.28190v1 Announce Type: new
Abstract: Embedding benchmarks like MTEB report a single score per model, implicitly treating robustness as a static, scalar property. We argue that embedding robustness is multidimensional, since models respond differently to different types of variation, and requires dynamic evaluation to expose failures hidden by static benchmarks. We introduce the Harder Text Embedding Benchmark (HTEB), a dynamic evaluation framework that challenges model robustness along three practically interpretable axes (Lexical/Stylistic, Length and Language) by stochastically transforming inputs at evaluation time with an LLM. Evaluating 16 open-weight embedding models on 32 datasets covering 42 languages under transformations validated by 4,800 human ratings on an English subsample, we find three patterns: (1) Models exhibit specific, partly decoupled robustness profiles across axes. (2) Across three model families, scale increases absolute scores but does not close the gap between original and transformed evaluations. Here, scaling tends to improve specifically the Language axis. (3) English datasets are more sensitive to HTEB transformations than multilingual datasets. This demonstrates that HTEB identifies strengths and weaknesses of models along deployment-relevant axes, challenging current embedding benchmarks and arguing for multidimensional, dynamic robustness evaluation.
arXiv:2605.28192v1 Announce Type: new
Abstract: Multi-hop audio-visual reasoning remains challenging for Omni-LLMs, as relevant evidence is often sparse, temporally dispersed, and distributed across both audio and visual streams. Existing benchmarks provide limited investigation of this setting, typically involving only a limited number of modalities, relevant temporal segments, or reasoning steps. In this work, we introduce MOV-Bench, a benchmark containing 519 carefully curated questions that require multi-hop reasoning over temporally dispersed audio-visual evidence. Evaluations on MOV-Bench reveal that current Omni-LLMs still struggle with multi-hop cross-modal reasoning. To address this challenge, we further propose AOP-Agent, an efficient agentic framework built on open-source Omni-LLMs for active omni-modal perception. By combining a hierarchical omni-modal memory with a collaborative observe-reflect-replan loop, AOP-Agent enables open-source Omni-LLMs to perform active perception without additional training or proprietary models. Experiments on MOV-Bench and OmniVideoBench demonstrate that AOP-Agent consistently improves reasoning performance, with particularly notable gains on long videos and reasoning-intensive questions.
arXiv:2605.28222v1 Announce Type: new
Abstract: We study quality-latency-resource trade-offs in a documentation-grounded retrieval-augmented generation (RAG) system that uses Low-Rank Adaptation (LoRA) of the generator. We build a manually verified benchmark of 5,144 question-answer pairs over the official Kubernetes documentation and combine it with a fixed hybrid-retrieval pipeline (BGE-M3 dense, BGE-M3 native sparse, Reciprocal Rank Fusion, cross-encoder reranking). Over this benchmark we ablate 20 LoRA configurations on Llama-3.2-3B-Instruct and Llama-3.1-8B-Instruct across rank and target-module choices, and evaluate each on token-level F1, LLM-judged groundedness and correctness (pass@4), inference latency, inference memory, and training cost, all reported with bootstrap 95% confidence intervals. Pareto analysis shows that LoRA adapters acting only on the q and v attention projections consistently dominate the front, while the 3B/8B choice mainly defines operating regime. A param-matched control comparison further indicates that the q/v advantage is structural rather than purely parametric. The benchmark, selected adapters, and code are available at https://github.com/EugPal/rag-lora-tradeoffs.
arXiv:2605.28224v1 Announce Type: new
Abstract: Multi-trajectory inference for tool-use LLM agents - generating multiple reasoning attempts and selecting among them - benefits from transferring knowledge across attempts so that later ones avoid the pitfalls of earlier ones. Existing cross-trajectory memory methods (trajectory-level reflection, atomic fact extraction, raw observation injection) are each evaluated under a single inference strategy on a single task, making it unclear whether reported gains reflect properties of the memory abstraction or of the inference method. We propose a unified framework that decomposes memory along two axes -- the scope of transfer (within an expansion vs. across trajectories) and the abstraction of the transferred content -- and evaluate four methods under three inference strategies (best-of-N, beam search, MCTS) on four tool-use benchmarks spanning SQL, knowledge-graph, and CLI environments, in a verifier-free setting that matches the deployment regime of practical agents. The experiment matrix identifies the inference method as a confound: the same memory method produces statistically distinct results under different inference strategies on the same examples. Reflection reaches significance only under MCTS (not under best-of-N); within-expansion injection (conditioning each candidate on prior siblings' outcomes) helps only diversity-starved beam search; and atomic fact extraction is accuracy-neutral but shortens trajectories by 19-26% on tasks with reusable environmental structure.
arXiv:2605.28227v1 Announce Type: new
Abstract: Speech translation models are increasingly capable of preserving speech-specific information (e.g., speaker gender, prosody, and emphasis), yet evaluation metrics remain blind to such phenomena. We meta-evaluate both text- and speech-based quality estimation metrics on two contrastive datasets targeting gender agreement and prosody, and find that both fall short, even when given direct access to the speech signal. We then train SpeechCOMET, a family of quality estimation models with speech encoders, and evaluate a state-of-the-art SpeechLLM as a judge. Both match or exceed text-based COMET on standard quality estimation, but neither consistently assesses speech-specific phenomena. We identify three causes: (1) speech-specific features are not reliably preserved in current encoders, (2) models tend to ignore the speech source signal, and (3) quality estimation training data contains too few relevant examples. We release all models and code, and argue that progress requires dedicated speech-specific training data and models that genuinely condition on speech.
arXiv:2605.28236v1 Announce Type: new
Abstract: Symmetry governs nonlinear interactions in condensed matter systems, particularly in high-harmonic generation (HHG), the interplay between the driving field and crystal symmetries dictate the properties of the emitted harmonics. A central open question is how quantum fluctuations of light modify these symmetry-imposed selection rules in solid state systems. Here, we address this by studying the nonlinear response of graphene and Molybdenum disulfide (MoS$_2$) to circular polarized quantum light, where both materials with distinct rotational symmetries and corresponding classical selection rules. We show that the quantum fluctuations break the dynamical symmetry of the driving field while preserving the crystal symmetry, which enables the generation of classically forbidden harmonics by breaking the corresponding selection rules. These results establish quantum states of light as a new degree of control over harmonic generation in solids, opening routes toward all-optical symmetry engineering of the quantum optical harmonic properties towards attosecond pulse generation.
arXiv:2605.28238v1 Announce Type: new
Abstract: Enforcing universal symmetries in machine learning (ML) models is a common strategy to mitigate data scarcity. We show that exploiting exact, as well as approximate, label symmetries can benefit scaling laws. We illustrate the idea for the s, p, d orbital densities of the electron in the hydrogen atom, for the three vibrational normal modes of the water molecule, as well as its full 3D potential energy hypersurface. Resulting ML models of electron density and potential energies exhibit superior learning curves, demonstrating improved generalization efficiency. When label symmetries are not exact, the same principles govern the observed learning behavior -- up to the convergence floors set by the degree to which the symmetry is approximate. For convex wells in the molecular potential energy surface, a Hessian-based correction suppresses the leading symmetry-breaking error in augmented labels.
arXiv:2605.28687v1 Announce Type: new
Abstract: Background: Infant cry acoustics provide a promising window into early neurodevelopment and may serve as scalable biomarkers for neurodevelopmental disorders. However, conventional microphone-based recordings are highly susceptible to environmental noise and raise privacy concerns in real-world clinical settings. Chest-surface accelerometers may offer a robust alternative by capturing vibrations directly from the larynx.
Methods: We evaluated the validity of a chest-mounted accelerometer (ACC) for infant cry analysis by comparing acoustic features derived from ACC and simultaneously recorded microphone (MIC) signals during routine vaccination visits. The final sample included 85 infants (41 at 4 months; 44 at 12 months) from a diverse pediatric population. Seven vocal measures were extracted from both modalities, including fundamental frequency (F0), jitter, shimmer, cepstral peak prominence (CPP), and harmonics-to-noise ratio (HNR). Agreement and consistency between modalities was assessed using intraclass correlation coefficients (ICCs).
Results: F0 demonstrated excellent agreement between ACC and MIC recordings (ICC > 0.94). Jitter measures also showed good-to-excellent agreement, while CPP demonstrated moderate agreement. Shimmer and HNR showed lower absolute agreement and systematic bias between modalities, reflecting possible differences in signal transmission and noise sensitivity.
Conclusion: In summary, chest-surface accelerometers can reliably capture several clinically relevant acoustic features of infant cry, particularly temporal measures of F0 and jitter. This approach offers a noise-robust and privacy-preserving alternative to microphone-based recordings, supporting its potential use in scalable clinical and developmental research applications.
arXiv:2605.27413v1 Announce Type: cross
Abstract: Proteins perform their biological functions through three-dimensional structures encoded by amino acid sequences, and ligand-binding protein co-design requires models that generate sequence-structure compatible proteins under explicit ligand constraints. Although continuous diffusion and flow-based models support ligand-aware design in coordinate or latent spaces, existing discrete diffusion protein language models mainly operate over sequence or structure tokens without direct small-molecule conditioning. We introduce \textbf{ProtLiD$^2$}, a \textbf{Prot}ein \textbf{L}igand-conditioned \textbf{D}iscrete \textbf{D}iffusion model for protein sequence-structure co-design. ProtLiD$^2$ jointly generates amino-acid sequence and discrete structure tokens while incorporating ligand chemical and geometric information through geometry-aware cross-attention. Trained on over one million ligand-protein complexes, ProtLiD$^2$ extends masked discrete diffusion to ligand-aware functional protein design. We further propose maximum confidence-margin guided ReMask decoding, an inference-time self-correction strategy that retains confident predictions and remasks uncertain tokens. ProtLiD$^2$ improves global fold confidence over Complexa in whole-protein design, increasing TM-score from 0.672 to 0.802 and pLDDT from 64.55 to 73.00. In pocket co-design, ProtLiD$^2$ reduces active-site BB-RMSD from 3.46/3.40{\AA} for FAIR/PocketGen to 1.97{\AA}, and improves ligand-aware pass rates over PocketGen from 14.86% to 59.73% and from 6.08% to 23.49% under stricter docking thresholds. These results support ligand-conditioned discrete diffusion as an effective token-space framework for functional protein co-design. Code will be available at https://github.com/auroua/ProtLiD.
arXiv:2605.27429v1 Announce Type: new
Abstract: Industrial video-on-demand (VOD) recommenders need richer content understanding, but LLM-as-reranker designs repeat prompt construction, token generation, model invocation, output parsing, and fallback handling for each request. In high-volume latency-sensitive services, these request-time operations complicate throughput planning, tail-latency control, capacity isolation, and predictable operation. This paper presents Ocean4Rec, a reranking layer that uses an LLM only offline to materialize item OCEAN profiles from content metadata. Items are mapped into Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism scores, while user profiles are built by time-decayed aggregation of recently clicked and deep-linked items in the same five-dimensional space. At request time, Ocean4Rec joins precomputed item profiles, user profiles, base recommender scores, and catalog recency, then performs numeric reranking without an LLM call. On anonymized Samsung Smart TV VOD logs, same-candidate Top1000 temporal-holdout offline evaluations show that Ocean4Rec improves NDCG@20 over a stronger non-OCEAN Base+Recency ordering by 7.6% for an NCF generator and 61.5% for a LightGCN generator. HR@20 is inconclusive for NCF and improves by 67.3% for LightGCN, reflecting sparse exact-item replay labels and the strength of recency as an industrial baseline. The result should be read as offline replay evidence for a bounded auxiliary content-taste feature that preserves the deployability advantage of a request-time-LLM-free serving path.
arXiv:2605.27432v1 Announce Type: new
Abstract: Retrieval-augmented generation (RAG) has emerged as a paradigm for grounding large language models in external knowledge, yet most existing RAG systems assume centralized knowledge access and ample computation. These assumptions break down in edge environments, where knowledge is fragmented across devices, raw data cannot be shared, and repeated LLM calls are prohibitively expensive. We propose FD-RAG, a federated dual-system RAG framework that decouples lightweight memory access from on-demand LLM reasoning for decentralized deployment. Specifically, FD-RAG learns semantic-aware adaptive hypergraphs over local corpora and distills them into compact QA memories. At inference time, it answers well-covered queries via direct memory matching and invokes LLM-based reasoning only when necessary, while tracing retrieved memories to hypergraph-grounded evidence. To mitigate cross-device knowledge fragmentation, FD-RAG aggregates anonymized memories across devices without exposing raw documents. Experiments on QA benchmarks show that FD-RAG improves accuracy by up to 7.8\% while reducing latency by 8.4$\times$ compared with strong local and federated baselines. We also provide theoretical analysis establishing an $\mathcal{O}(1/\epsilon^{2})$ convergence rate for the proposed hypergraph learning, supporting its tractable deployment in edge settings.
arXiv:2605.27439v1 Announce Type: new
Abstract: AI assistants like ChatGPT and Claude are recommendation engines, not search engines: they answer commercial queries by directly nominating brands rather than returning a list of links. Marketing to AI is therefore a broader problem than "show up in search" -- positioning, content, and product fit matter as much as discoverability. We audit ~37,000 production runs across four model configurations and 215 commercially-framed prompts spanning 19 sectors, evaluated against a 533-brand reference catalog stratified into five prominence tiers (L1 category leaders to L5 regional players) sourced from external authority lists. The ladder proxies a brand's awareness footprint within its sector, not revenue or market share. The failure mode differs sharply by tier. L1 brands appear in nearly every relevant retrieval but win only 25-41% of the recommendation slots they reach -- the leverage is differentiation, not visibility. L2 challengers carry the highest conversion rates of any tier (37-52%) but lose to persona-mediated substitution on the Anthropic models. L3 mid-market brands are the inflection level: aggregate coverage drops to 88%, conversion to 34-40%, and persona effects peak. L4 specialists and L5 regional players face catastrophic invisibility -- 48-52% never surface in any of the 37,000 runs. No uniform optimization recipe wins; the right marketing investment depends on where the brand sits on the prominence ladder.