How Reliable Are Semantic-ID Tokenizer Comparisons in Generative Recommendation?

In Semantic-ID (SID) based generative recommendation, each item is represented as a sequence of discrete codes, and an autoregressive model is trained to generate the SID sequence of the next item; top-K performance is then measured by checking whether the SID sequence of the target item appears among the generated sequences. This evaluation protocol equates SID-level matching with item-level recommendation, an equivalence that holds only when every SID sequence maps to a single item. We show this assumption breaks down in practice: because tokenizers compress item features into a code space, semantically similar but collaboratively distinct items are frequently assigned the same SID sequence. Across four datasets and five representative tokenizers, the fraction of items involved in such collisions reaches 30.5%, so matching a shared SID sequence identifies only a collision group rather than the target item. Consequently, SID-level metrics overestimate item-level performance (Hit@10 is inflated by up to 103.36%), and the inflation grows with the collision rate. To support faithful comparison, we develop collision-aware item-level metrics computed directly from generated SID sequences, together with a post-tokenizer procedure that reassigns last-level SIDs at minimum cost to obtain a collision-free assignment for any existing tokenizer. Our results indicate that SID-level rankings in prior work should be interpreted with caution, and that reliable tokenizer evaluation requires either item-level correction or collision-free SID assignments.

Choosing Online Experiment Designs under Interference in Ads, Recommendations, and Member-Experience Systems

Online experiments in ads, recommendation, and member-experience systems are often planned before the dominant interference mechanism is known. A treatment may propagate through budgets, inventory, producer exposure, graph spillovers, or temporal carryover, making the randomization design itself a statistical decision. We formulate this problem as robust design selection over uncertain exposure mechanisms. Given a finite catalog of six implementable designs, the selector compares each design by worst-case planning risk over an ambiguity set. The risk combines exposure bias, assignment-unit variance, minimum detectable effect, contamination or carryover, operational cost, and estimand mismatch. For theoretical justification, the paper develops a geometry-aware guarantee, stating that design bias is bounded by Wasserstein distance to the launch exposure distribution, and this penalty is minimax tight under Lipschitz exposure response. We also prove finite-catalog approximation and a robust selector theorem with excess-risk control, exact recovery under separation, and certified shortlists when the risk surface is flat. Empirically, the same selector gives different recommendations across samples from public datasets. It selects user-randomization on Criteo ads with dimensionless robust risk 1.295, switchbacks on Open Bandit-bts/men with risk 2.105, and cluster-randomization on KuaiRand with risk 2.240. The Open Bandit case stresses known but uneven logging support, with propensities from 0.00006 to 0.594 and a 5.17% IPS effective-sample share. Overall, the paper contributes an interference-aware experiment design framework based on mechanism-robust design decisions, where the output is either a justified design choice or an uncertainty shortlist.

LLM Agent Based Renewable Energy Forecasting Using Edge and IoT Data A Review of Solar Wind Weather and Grid Aware Decision Support

Reliable forecasting of renewable energy generation is a foundational requirement for grid stability energy trading battery scheduling and carbon aware operational planning Solar and wind resources are inherently intermittent their output fluctuates with cloud cover wind speed atmospheric turbulence seasonal patterns and local terrain The proliferation of IoT and edge devices spanning smart meters inverters anemometers pyranometers weather stations and grid interface sensors has created an unprecedented volume of real time operational data that conventional forecasting pipelines are ill equipped to exploit fully This review investigates how large language model LLM agents can enhance renewable energy forecasting by integrating heterogeneous sensor streams weather API data historical generation records grid constraints and contextual reasoning into unified decision support workflows We survey classical forecasting methods statistical time series models deep learning architectures physics hybrid approaches and emerging LLM agent frameworks for explanation uncertainty communication and operator guidance A six layer taxonomy is proposed covering data acquisition preprocessing feature engineering model inference uncertainty estimation and natural language reporting The review identifies twelve open challenges spanning real time deployment model drift under distribution shift uncertainty quantification hallucination control in LLM agents interoperability of edge hardware and integration with energy management systems The paper concludes by recommending a research agenda centred on open benchmarks physics informed LLM grounding and federated forecasting architectures

Context Features Are Cheap: Rank-Aware Decomposition for Efficient Feature Interaction in Recommender Systems

Modern industrial recommender systems use a deep ranking model to score N candidates against the same user and context features. Standard implementations broadcast context features early in the forward pass, redundantly computing context-only operations N times per request. We present a rank-aware decomposition applicable to the dominant interaction mechanisms in modern recommender architectures-Factorization Machine (FM) pairwise products, Deep Cross Network (DCNv2) cross layers, self-attention, and fully connected (FC) projection layers-built on a single algebraic principle: any linear or bilinear operation over a rank-partitioned input admits an exact block decomposition that moves context-only computation from once-per-candidate to once-per-request, identity-equivalent to the original model. Closed-form analysis and controlled ablation verify that savings scale quadratically with the number of context features. Applied to a production DLRM-style ranker without any architectural change, the decomposition increases per-pod throughput by 87.5% (a 47% reduction in peak pod count) at identical model predictions. The identity-equivalent decomposition applies only at the first layer of cross networks and self-attention, since each layer mixes ranks in its output. To extend savings across depth, we further introduce rDCN, an architectural variant of DCNv2 that maintains rank discipline across depth and matches DCNv2 accuracy within training noise at 67% fewer total FLOPs, and sketch an analogous architectural variant for self-attention.

Active Learning for Stochastic Contextual Linear Bandits

A key goal in stochastic contextual linear bandits is to efficiently learn a near-optimal policy. Prior algorithms for this problem learn a policy by strategically sampling actions but naively (passively) sampling contexts from the underlying context distribution. However, in many practical scenarios -- including online content recommendation, survey research, and clinical trials -- practitioners can actively sample or recruit contexts based on prior knowledge of the context distribution. Despite this potential for active learning, the role of strategic context sampling in stochastic contextual linear bandits is underexplored. We propose an algorithm that learns a near-optimal policy by strategically sampling rewards of context-action pairs. We prove instance-dependent theoretical guarantees demonstrating that our active context sampling strategy can improve over the minimax rate by up to a factor of $\sqrt{d}$, where $d$ is the linear dimension. We show empirically that our algorithm reduces the number of samples needed to learn a near-optimal policy, in tasks such as warfarin dose prediction and joke recommendation.

First, do no harm: Breaking suicidogenic echo chambers in media recommendation

Recommender systems generally optimises user engagement, but this approach is dangerous in mental health contexts. When vulnerable users show signs of suicidal ideation, standard algorithms often trap them in echo chambers of harmful content, worsening their psychological state. In response, we introduce RankAid, a re-ranking method that prioritises clinical safety alongside predictive relevance. It works as an add-on layer to existing models: it penalises risky items and boosts therapeutic content depending on the user's current level of vulnerability. We evaluated this approach using the MovieLens 1M dataset, where items were semantically annotated for clinical risk and therapeutic value using large language models. Our simulations show that our algorithm successfully blocks the recommendation of harmful content during crisis peaks, actively reshaping the feed to support emotional de-escalation. Furthermore, this safety intervention only causes a controlled, acceptable drop in standard accuracy metrics like NDCG. By using asymmetric hyperparameters, RankAid also gives system administrators the flexibility to tune the severity of the intervention based on specific clinical guidelines.

Meta-Modal Agent: Sequential Evidence Routing for Missing-Modality Candidate Reranking

Missing modalities cause severe failures in multimodal recommender systems. User histories, item text, and visual evidence are frequently absent during cold-start scenarios, exactly when recommendation quality matters most. Existing approaches recover absent signals through imputation, feature propagation, or generative reconstruction, but these strategies can inject unsupported evidence when the surviving signals are weak. We introduce the Meta-Modal Agent (MMA), a large language model based candidate-pool reranker that treats missingness as a sequential evidence-routing problem. MMA is trained with balanced missingness-task reinforcement learning over masked-modality episodes and is evaluated in two variants: MMA-Auto, which uses only automated text, image, and graph tools, and MMA-Interactive, which additionally permits clarification questions grounded in surviving modalities as an upper-bound diagnostic. MMA operates after a first-stage retriever has produced a candidate pool; it scores those candidates rather than retrieving items from the full catalog. Final reranking fuses MMA scores with first-stage retrieval scores selected on validation data. Our evaluation is organized around four evidence checks required for a robust missing-modality claim: oracle-free one-observed-modality availability (OOMA) robustness, per-modality OOMA breakdowns, fixed-pool full-catalog reranking, and a deterministic-router mechanism control. MMA-Auto improves target-positive OOMA NDCG@10 by 4.0% and fixed-pool full-catalog reranking NDCG@10 by 12.7% over the strongest non-interactive baseline. RuleRouter-Fuse, which uses the same tools and fusion rule without learned policy updates, underperforms MMA-Auto, supporting learned routing beyond deterministic tool fusion. MMA-Interactive adds a 4.1% upper-bound gain when clarification is available.

RAGe: A Retrieval-Augmented Generation Evaluation Framework

Deploying Large Language Model (LLM) applications, particularly those relying on Retrieval-Augmented Generation (RAG), remains challenging due to high computational demands, outdated knowledge bases, and the need to manually select optimal pipeline components. In this work, we propose a modular framework for benchmarking and guiding the efficient development of RAG applications by focusing on resource telemetry and component recommendation, suggesting the best components for a domain-specific dataset. Our approach leverages core techniques in LLM applications, including document chunking, vector databases, embedding models, and retrievers, to evaluate trade-offs among accuracy, efficiency, and scalability. By directly correlating retrieval and generation quality with underlying hardware constraints, RAGe supports researchers to identify the most effective, domain-specific RAG setups for their specific operational needs, facilitating rapid prototyping even on consumer-grade hardware.

When Mean CE Fails: Median CE Can Better Track Language Model Quality

Mean cross-entropy is the standard validation metric for language models, but it can fail to track model quality during training. We examine this in two common scenarios. First, in Qwen2.5-1.5B SFT on synthetic fact-learning, we find that mean CE rises substantially after the initial learning phase while held-out fact-recall accuracy remains near its peak. Second, we find that in top-K distillation on TinyStories, decreasing K improves median CE while worsening mean CE; the Top-5 student attains the highest LLM-judge score and crosses below its teacher on median CE, despite having the worst mean CE. In both cases, median CE correlates much more closely with task performance than does mean CE. Analyzing how bulk and tail percentile CE move during training reveals that training reshapes the empirical per-token CE distribution. In top-K distillation, smaller K yields a distribution with more mass at both extremes, decreasing the median and increasing the mean. In Qwen SFT, the bulk saturates quickly while the tail extends in the latter half of training. In both, the task-evaluation metric appears more sensitive to the bulk than to the tail. Practically, we recommend reporting a small set of percentile CE summaries alongside the mean, and using concordance among them as a tool to keep track of distribution reshaping, as well as a low-cost diagnostic for when mean and median CE disagree on model selection.

A Signal Extraction Approach for Remote Heart Rate Variability Assessment Using Proxy Measure in a Driving Simulator

This study evaluates remote Photopletismography (rPPG) algorithms, Spatial Subspace Rotation (2SR), Chrominance-based method (CHROM), Plane-Orthogonal-to-Skin (POS), and Principal Component Analysis (PCA), applied to selected superpixel-based facial regions (with target counts of 10 and 20 regions) for monitoring in a driving simulator. Two novel peak enhancement approaches, based on the Lp norm and Fractional-Order Derivative (FOD), are introduced to enable robust Heart Rate Variability (HRV) estimation. A signal-to-noise ratio-based quality assessment of 20 s segments serves as a data cleaning mechanism to mitigate motion artifacts inherent to dynamic recording conditions. In a sample of 29 participants recorded during baseline and driving simulation conditions, Pulse Rate (PR) is calculated with clinically acceptable accuracy across configurations (validated against simultaneous Electrocardiography (ECG) recordings), achieving the lowest Mean Absolute Error (MAE) of 1.92 bpm (sd = 1.72) using 2SR with FOD and 20 superpixel regions. The best-case MAE reached 0.061 s for Standard Deviation of Normal-to-Normal intervals (SDNN) and 0.081 s for Root Mean Square of Successive Differences (RMSSD), with inter-beat interval detection yielding an F1 score of 0.93. Optimal parameters clustered around p = 6-7 for Lp norm and fractional orders of 1.0-1.4. All rPPG-derived parameters reproduced the statistical structure of the reference ECG across conditions and configurations. Caution is advised when using FOD due to slow changes in the rPPG waveform. Overall, 2SR is recommended for PR, while CHROM for HRV estimation, using Lp norm with 20 superpixels, providing clear methodological guidance for rPPG monitoring in driving simulators