Papers

Accurate job runtime prediction can improve scheduling-aware resource management in grid and distributed computing environments, but prediction models must be evaluated under realistic deployment constraints. This paper revisits CPU burst time prediction on the GWA-T-4 AuverGrid workload trace and reformulates it as leakage-safe pre-execution job runtime prediction. We define the target as job-level runtime, use only submission-time attributes, exclude post-execution variables, and evaluate models under temporal and cold-start settings rather than relying only on random cross-validation. We compare standard regressors, chronological historical baselines, categorical encoding strategies, and CatBoost with native categorical handling. We further add temporal hyperparameter tuning, runtime predictability analysis, feature ablation, error analysis by job length, and a minimal scheduling simulation. After temporal-validation tuning, CatBoost achieves the strongest deployment-oriented result with R^2=0.239, MAE=27,019, RMSE=46,587, and LogMAE=2.646 on the held-out temporal test set. A single-server simulation over all 69,523 held-out temporal test jobs shows that prediction-informed SJF reduces average waiting time by 50.92% relative to FCFS. The results show that random-split evaluation overestimates performance, categorical-native boosting improves temporal generalization, and long-job underestimation remains a scheduler-relevant challenge.

Negation remains a longstanding challenge for both language models (LMs) and large language models (LLMs). Prior work mainly focuses on a small set of high-frequency single-word negation cues, such as not and never, with limited exploration of broader negation types and modern LLMs. To address this gap, we construct NegCue, a large-scale dataset containing over 1.8M samples spanning single-word, multi-word, and affixal negation with more than 200 unique cues. We further pre-train both encoder-only LMs and LLMs on NegCue to investigate how different negation types affect negation understanding. Experiments on five downstream benchmarks show that negation types contribute unevenly to performance gains under the same training scale. In particular, affixal negation yields the largest improvements, while the gains from the commonly studied single-word negation remain modest. Moreover, our results demonstrate that further pre-training improves negation understanding for both LMs and LLMs.

We introduce LayerRoute, a parameter-efficient method for adaptive transformer layer-skipping that combines per-layer hard-gated routing (trained via a straight-through estimator) with joint LoRA fine-tuning. LayerRoute augments each of the 24 transformer blocks in Qwen2.5-0.5B-Instruct with a lightweight per-layer router (~21.5K parameters) and LoRA adapters (rank 8, ~1.08M parameters), training both jointly under a gate-regularized language-modeling objective. Across 10 independently-seeded training runs, LayerRoute converges to an identical skip-pattern structure in every run - a consistent set of 9 middle layers (8-16) becomes skip-eligible in all 10 seeds - and delivers genuine, verified wallclock speedup in every run (1.02x-1.06x, mean 1.04x). Quality is preserved or improved in every configuration tested: joint LoRA adaptation yields a perplexity improvement over the unmodified backbone in all 10 seeds (mean delta = -1.16 and -1.11 across the two evaluation splits used). We further verify the router performs genuine, non-trivial per-input computation: gate decisions in skip-eligible layers change the actual skip/run outcome for 87-100% of held-out samples, confirming real input-dependent routing rather than a fixed pruning pattern. LayerRoute trains in under 7 minutes on a single A100 and adds negligible overhead beyond the routing decision itself. We report our full reproducibility methodology, including a systematic diagnostic investigation into what determines the router's per-input decisions, as part of this work.

Fine-tuning instruct models often improves target performance while inducing behavioral drift from the reference model, which can degrade existing capabilities. Rather than treating this drift as an uncontrolled consequence of optimization, we specify a behavioral drift budget before optimization and ask how to boost the target-task performance within it. Locally, behavioral drift induces a shared geometry anchored at the reference model, with the drift budget defining a boundary within this space. In this space, drift determines distance from the reference, leaving update direction as the remaining degree of freedom. Fine-tuning updates can therefore be compared through their directional efficiency, naturally reformulating fine-tuning as a direction-selection problem. This reformulation makes a concrete prediction: changing the accessible directions can qualitatively alter the outcome of fine-tuning. We test this prediction in a stringent QA-only setting, where strong instruct models are fine-tuned only on final answers but must still generate multi-step reasoning at inference. Despite this mismatch, a coarse layer-selective probe reverses the failure of QA-only fine-tuning and reveals the existence of effective directions, with multiple neighboring configurations improving target performance while preserving reasoning and general capabilities. Across Qwen3-8B and Qwen3-14B, these directions substantially improve scientific reasoning and multilingual translation. Over more than 100 languages, the resulting models match or outperform dedicated translation systems and provide a stronger initialization for subsequent reinforcement learning. Our results suggest that fine-tuning is not just about how much a model changes, but how that change is spent. https://github.com/CONE-MT/DCO and https://huggingface.co/collections/LLaMAX/dco

Markov decision processes (MDPs) are used to support decision-making in conservation of biodiversity, but policies, even over small state spaces, can be difficult to interpret for conservation managers. K-MDP methods address this problem by building simpler MDPs with at most K abstract states. We show that the previously proposed A-K-MDP algorithm that relies on selecting a discretisation divisor using binary search can skip better abstract states. To fix this issue, we propose Windowed A-K-MDP, an algorithm that generates every distinct feasible partition induced within a declared divisor window and evaluates candidates until reaching the ideal value loss (J = 0) or exhausting the family of candidates. Across 33 K-MDP instances, Windowed improved 25 and tied 8.

AI agents can be interrupted while editing files, calling tools, or carrying out multi-step tasks. Restarting repeats completed work, but continuing from unverified or outdated progress can carry earlier errors forward. A saved state is not necessarily a suitable place to resume. We introduce recoverability as a system primitive that makes reuse an explicit decision: select a supported starting point and a permitted recovery action, or withhold automatic continuation. Its behavioral contract binds that choice to supporting evidence, execution, and independent checks. A reference architecture connects persistence, validation, and control, with complementary runtime instances testing distinct responsibilities. Four deterministic and 20 paired file challenges demonstrate that accurate restoration and successful completion can conceal disallowed starting points. Progress controls attribute retained work to shared restoration. Event-time tests show that permission must also constrain the action, and that independently held policy evidence can expose violations even after an effect occurs. These findings establish why recovery decisions need their own evaluation, beyond restored bytes and final task success. Within supplied policies and a declared trust model, the contribution is a common, testable interface for retaining justified progress and making the conditions for its reuse explicit and enforceable.

Event linking associates event mentions in text with entries in a knowledge base (KB), or identifies them as out-of-KB events. Although existing methods use different architectures, candidate event acquisition can still be weakened by short ambiguous mentions, noisy arguments, and evidence that is unevenly useful for retrieval. We present MACE, a Multi-Agent Candidate Event acquisition method that refines event structure before linking. MACE uses evidence-specialized LLM agents to acquire time, location, participant, and event-type evidence, exposes intermediate queries to candidate-event lookup tools, and lets a coordinator revise the evidence set before final candidate construction. Experiments on two event linking benchmarks show that adding MACE to different event linking models consistently improves accuracy. These results show that MACE improves event linking through better candidate event acquisition without modifying the event linking model.

Geospatial agents are increasingly expected to support recurring and evolving analytical tasks rather than execute isolated workflows. In such settings, effective agents must distill prior execution experience into reusable geospatial procedural knowledge to guide future planning and tool use. However, existing memory-augmented paradigms struggle to summarize both long-horizon tool-chain orchestration experience and tool-level invocation constraints in geospatial analysis, while directly relying on LLM self-reflection to update experience often leads to misattribution and unreliable revisions. To address these challenges, we propose GeoSkill, an experience-driven hierarchical skill learning framework for geospatial agents. GeoSkill comprises two core components: (i) a Hierarchical Skill Bank (HSB), consisting of a Planning Skill Bank and a Tool Skill Bank, which respectively distill high-level task-planning experience and tool usage constraints, enabling structured representation and cross-task reuse of historical execution experience; and (ii) a Collaborative Trace-driven Skill Revision (CTSR) mechanism, where Judge, Critic, and Refiner collaboratively perform error identification, skill-level defect localization, and targeted modification, preventing misattributed and unreliable revisions from polluting the skill bank. GeoSkill learns and validates skills from historical executions during development, and freezes the skill bank for retrieval-only guidance on unseen tasks during deployment. Extensive experiments on EarthBench and ThinkGeo demonstrate that GeoSkill effectively transforms historical execution experience into reusable hierarchical skills, improving both end-to-end task accuracy and tool-execution reliability in geospatial tasks.

Knowledge graphs are increasingly distributed across autonomous organizations that share an entity space but own disjoint subsets of relations, forming a vertical partition. Answering a multi-hop query may require combining facts from several silos, making the partitioning strategy a key data management decision that affects communication, indexing, load balance, and query latency. However, the costs associated with different partitioning strategies remain insufficiently studied. We formalize vertical partitioning as a design space and compare four strategies: semantic domain grouping, frequency-balanced partitioning, co-occurrence graph-cut partitioning, and random partitioning. We evaluate them using five metrics: communication cost, candidate index size, cross-silo path length, load balance, and end-to-end query latency. Three of the five prove to be determined by the graph and the silo count rather than by the partition, which reduces the design problem to two conflicting axes, cross-silo path length and load balance. Experiments on MetaQA and PathQuestion use a fixed federated knowledge graph question-answering architecture based on TransE embeddings and a frozen BERT encoder across three silo configurations. By keeping the learning model unchanged, we isolate the effect of partitioning and show that the trade-off between locality and balance holds only where each silo can hold several relations, weakening as the number of silos increases. The study provides practical guidance for deployments constrained by cross-silo reasoning or by silo load.

Knowledge graph question answering usually assumes that one system can reach the whole graph. In practice, facts are often held by organizations that share entity identifiers but own disjoint relation types, so no single party sees a complete reasoning chain. This poster presents the empirical findings of FedV-KGQA on multi-hop question answering over such vertically partitioned graphs. Each silo enriches its local graph and trains a knowledge graph embedding on its own triples. A server then concatenates the silo-specific entity views, anchors the projected question at the topic entity, and ranks candidates by similarity. Raw triples and relation embeddings never leave a silo. Comparing the FedV-KGQA experiments with one another yields three results. First, federated fusion recovers most of the centralized accuracy, while a single silo recovers little. Second, anchoring and enrichment matter more than the choice of embedding model. Third, the cheapest encoder depends on the target accuracy rather than on parameter count. This poster paper contributes that cross-experiment comparison, four design lessons drawn from it, and an interactive prototype that runs real inference and traces the full pipeline, per question, on released checkpoints.

Traditional recommender systems are typically trained to predict what item users will interact with next, but not why. However, offering personalized evidence for why a user might like the predicted item is an important way to enhance the service and to raise the likelihood that the user will be genuinely interested in the recommendation. This service can be delivered by integrating a frontier-model call into the member-facing pipeline, but it will add extra cost and latency. In this paper, we train a recommender LLM to generate personalized explanations for its reccomendation, based on the user's watching history at a large video streaming service. We impose two requirements on the generated explanation: it must be faithful to the elements of the shows it links, and it must be strictly non-harmful to the user. To this end, we first train two LLM-judge reward models covering three specific criteria, and propose constrained GRPO to incorporate these different criteria. On a held-out real-world testing set, our fine-tuned model improves the all-three-criteria PASS rate rises from 0.649 to 0.956 under our own judges and from 0.677 to 0.931 under an independent judge, where as the frontier generator performs similar to the untuned recommender baseline. We conduct further experiments to show that the model's language and recommendation abilities remain unchanged. Based on these results, we conclude that an LLM-based recommender can be fine-tuned on other complex tasks without compromising its original recommendation performance, thus provide insights for further agentic user interface powered by a single model.

Solar energy decision support is fragmented across dashboards that provide data without explanation, research papers are slow to parse, and general-purpose language models are not solar domain specific and answer without evidence. This paper introduces Solar Intelligence, a hybrid retrieval-augmented framework that unifies structured solar analytics, evidence-grounded scientific question answering, and machine learning forecasting in one system. The platform integrates daily NASA POWER solar and meteorological data, Biosphere 2 ground-sensor readings, and a curated corpus of research papers and institutional reports. Structured queries use DuckDB SQL; scientific questions are answered by a hybrid retriever that fuses BM25 and ChromaDB dense embeddings via Reciprocal Rank Fusion, with responses grounded through a language model (llama3.2:3b). An Extreme Gradient Boosting (XGBoost) model produces daily forecasts of irradiance, temperature, and wind speed. The system is exposed via FastAPI, Streamlit, and an MCP server, so it can be used as an application, an API service, or an agent tool - by students, researchers, and energy analysts.

Persistent agents need evaluations that distinguish identity facts they can recall from those they express and enact. We introduce PAI-Bench, a provider-neutral benchmark for fidelity to a versioned, update-governed identity contract. It separates recall, composition, behavioral enactment, resistance, persistence, lineage, and role-conditioned updates while keeping scoring oracles outside the target process. Two frozen campaigns cover sixteen synthetic profiles, thirty-two probes, and three independently initialized target configurations, yielding 1,536 retained responses. A judge-independent literal audit finds direct-parent identifiers in 48/48 atomic responses but only 1/48 implicit self-portraits. On eight profiles, explicit field cues increase joint presence of three identity identifiers from 0/8 to 7/8 under the same four-sentence instruction. A separate startup body-label substitution increases full-designation presence from 1/8 to 7/8 while parents remain absent. These contrasts reveal prompt-dependent component selection and component-specific sensitivity to startup cues in the tested deployments. Replaying identical factorial responses also yields a Claude headline mean 12.5 percentage points below Astra's, demonstrating evaluator sensitivity separately from target behavior. The studies use single target samples per condition, with post-hoc audits and follow-ups. PAI-Bench provides a reproducible evaluation protocol for measuring factual availability, identity expression, and behavioral enactment as distinct aspects of identity-contract fidelity.

Do AI assistants help believers reason about moral dilemmas consistently with their faith? We present FaithfulBench, the first benchmark to score AI counsel across traditions by how well it adheres to the user's professed faith. Scenarios are drawn from each tradition's most respected texts, with the faithful answer known and applied by the judges as the standard. We test five frontier models under three conditions: the AI does not know the user's tradition; it receives a one-line prompt identifying the user as a practicing adherent; or it receives a companion-counselor guide rooted in the tradition's sources. Two judges score the initial response and whether the model caves or holds when pressured toward the answer the user wants. When the tradition is unstated, models counsel from a secular therapeutic default and every model fails some believers. Naming the faith wins a faithful first answer but not steadfastness; the guide improves both.

Large Language Models (LLMs) are powerful zero-shot learners but remain prone to misalignment with human preferences, often producing biased, toxic, or otherwise harmful outputs. Existing alignment methods, while effective, are costly and tightly coupled to the model, limiting flexibility and scalability. We propose a modular correction framework that augments pretrained LLMs with Activated LoRA (aLoRA) adapters and a context-aware routing mechanism to eliminate harms from misaligned model responses. Our approach enables expert adapters to activate mid-sequence without invalidating the KV cache, allowing low-latency, targeted correction during generation. Each expert is trained to detect and mitigate specific harms, such as bias or toxicity. A learned router dynamically selects appropriate experts based on the models intermediate outputs. We demonstrate that our system improves alignment on standard safety benchmarks while preserving task performance, offering a lightweight and efficient path toward safer and more controllable LLM deployments.

Continual Learning Bench scores learning as short-horizon gain versus a reset baseline and finds naive full-context ICL strongest among the memories it tested. We treat ICL as one learning system and score it on a longer shared-world schedule. Steady-state learning is the gap versus baseline on a pre-set late window (last 40 of 174 BIRD-SQL formula-1 questions). We split the score into exploration efficiency (SQL probes), task reward (hits), and delivery cost (API dollars and context size). On gpt-5.6-luna, late probes fall from 4.6-5.6 to 0.95 while hits rise only modestly and ICL context grows to about 95k tokens with cost roughly doubling. Short-horizon gain understates the late probe saving and misses the cost inversion, so we find that unbounded ICL is a poor candidate for the learning mechanism.

Diagnosing rolling-element bearing faults from vibration is a canonical physical-sensing task and a widely used benchmark for domain adaptation under operating-condition shift. Accuracies above 99 percent are commonly reported, but under evaluation splits that place the same physical bearing in both training and test. We revisit the task under a held-out-bearing protocol, assigning every bearing unit entirely to either the training or the test set, and find that source-only transfer is far weaker than such numbers suggest: on a change of shaft speed it reaches only $0.36$, against a target-supervised ceiling of 0.97. We then study what governs transfer. Treating computed order tracking, a shaft-angle resampling that places fault frequencies at fixed shaft orders independent of running speed, as a controlled change of representation, we find that a Fourier Neural Operator raises source-only transfer from $0.36$ to $0.61$ on the speed shift, where the fault peaks move, while a convolutional network of matched feature dimension stays near chance in both representations. The representation also decides whether unsupervised alignment can work: with the same normalized RBF-MMD loss and no target labels, the operator reaches 0.71 in the frequency domain but 0.95 in the order domain, within 0.02 of the target-supervised ceiling and above $0.86$ on every held-out bearing fold. Once the representation is right, a small label budget adds little. These results indicate that, for this task, the input representation rather than the alignment method decides whether adaptation helps. A second dataset, whose held-out units are fault diameters rather than bearings, shows that the same protocol exposes failures that even a target-supervised model cannot avoid.

Modern design optimization frameworks aim first and foremost for models with the most accurate predictions without balancing computational overhead. It remains a reason why scaled architecture and multidisciplinary design optimization problems are difficult to address, even with sample-efficient Bayesian optimizers. In this paper, a metric quantifying the computational energy footprint is introduced within a Bayesian optimization framework to guide the parameter setting of a model towards configurations that balance both performance and frugality. The computer experiments highlighted existing tradeoffs between optimum convergence and the underlying energy footprint, and sometimes resulted in both a better-found optimum and lower energy consumption.

The purpose of this article is to highlight the central role of autonomous systems as the ultimate stage in the development of AI, to explain the underlying technical challenges that require a combination of connectionist AI and symbolic AI, and to integrate AI and systems engineering. We present a comprehensive framework for the design and evaluation of autonomous systems, based on a generic agent architecture that characterizes their behavior as the composition of cognitive functions organized around a long-term memory containing the agent's evolving knowledge. We address the challenges posed by the implementation of the fundamental features of the agent architecture, in particular the link between sensory data and structured data stored in memory, decision-making related to the achievement of the agent's goals and their planning, as well as the coordination of agents to combine individual and collective intelligence. We explain that agent trustworthiness, unlike that of traditional systems, is not limited to behavioral properties. It includes an essential dimension related to cognitive properties, the validity of which depends on how the agent uses its knowledge in decision-making. We present avenues for the development of methods for evaluating agent trustworthiness. We conclude with a critical assessment of the substantial gap between the aspirational vision of autonomous multi-agent systems and the current state of the art.

Evolutionary Ensemble Search (EES) constructs machine-learning procedures through expert-guided program evolution. A role-specialized council turns task evidence and experimental results into structured search directions. An orchestrator allocates these directions to execution specialists and an evolutionary engine. The engine selects measured parents, diagnoses their errors, and produces descendants through code mutation, structured pipeline edits, and crossover. Each child must execute and acquire its own validation evidence. Population archives retain useful alternatives, while compatible predictions compete in a validation-gated ensemble stage. Search adapts through parent-relative operator credit, session memory, and problem-indexed lessons retrieved across runs. We specify these mechanisms, distinguish their execution profiles, and define the contracts required to compare candidates as their computations change. A public MLE-bench Lite development ledger records medal-threshold artifacts on 19 of 22 tasks (86.36%), with best outcomes of 11 gold, five silver, and three bronze. The procedures span text, images, tables, audio, scientific geometry, and deterministic transformations. The campaign includes grade feedback between runs, external-source routes, and mixed confirmation procedures; its aggregate is an achieved development result, not a blind autonomous-agent success rate. The report contributes a concrete architecture for cumulative executable search and a versioned account of its cross-modal development outcomes.

Voice agents often need to collect names, addresses, identifiers, dates, and times exactly, yet end-to-end benchmarks obscure where capture fails. We introduce $\tau$-Elicitation, a 200-task voice benchmark spanning 10 entity types, controlled difficulty, caller realisms, and three environments. A matched text agent passes all tasks, but four voice configurations achieve robust exact success from 0.14 to 0.41. Agents increase verification for hard and unfamiliar entities and sometimes for incorrect captures, but not for their weakest caller voice; only 24 to 37 percent of verified errors are repaired. A scaffold that prescribes spelling, read-back, correction, and confirmation raises robust Pass$^3$ by 14 to 31 points, at a cost of 21 to 28 seconds per call. Realisms such as spelling variations and restarts do not detectably affect exact success; mispronunciation increases repair effort. These results identify strategy selection and successful recovery as the central bottlenecks in exact spoken entity collection.

Predicting potential dangers is core to safety. Forecasting the presence of other traffic agents is core to danger prediction. Occluded traffic agents challenge detection systems as they might become visible too late, leaving the autonomous vehicle too little time to identify, plan and act accordingly in a robust and safe way. Previous works proved that auditory perception, being omnidirectional and not constrained by a field-of-view, provides fundamental cues for early spotting of different road users, even when hidden by other vehicles or infrastructures. Yet, those contributions deal with scenarios with only one vehicle present, and they either identify the type of vehicle or estimate its direction of arrival. In this work, we move forward, and propose a multi-task system which, simultaneously, estimates the number of vehicles present, their type, and their direction of arrival. Our methodological contribution is a two-stage pipeline: a self-supervised pre-training stage inspired by the Joint- Embedding Predictive Architecture (JEPA) applied directly to multichannel raw waveforms, followed by supervised multi-task fine-tuning with a bidirectional LSTM and three classification heads. The pre-training stage trains the encoder without labels, pushing it to predict the latent representation of a future audio segment from its past context. To train and test our framework, since no suitable dataset was publicly available, we collected an ad-hoc one covering Non-Line-Of-Sight scenarios with multiple traffic agents simultaneously operating. Experimental evaluation shows that our method outperforms the state of the art; it also proves that our design choice allows the model to learn robust representations, which can be transferred to an unseen driving scenario, maintaining reasonable and stable performance.

Communication has become a bottleneck in distributed training and inference of large models. Overlapping communication with computation at the granularity of kernels, on separate streams, reduces only part of this communication cost. Fused kernels often have better performance by transmitting each output tile as soon as it is produced, but existing fused kernels are largely confined to a single NVLink domain. We present mKernel, a library of multi-GPU, multi-node fused kernels that overlap computation, intra-node NVLink communication, and inter-node RDMA at tile granularity. mKernel partitions the streaming multiprocessors (SMs) of a persistent kernel into compute and communication roles, and an on-GPU controller tunes the SM partition adaptively at run time, since the best SM partition varies with the kernel and the input shape. It structures data movement hierarchically so that data traversing the inter-node network is minimized. Finally, it drives the network from the GPU through a lightweight command queue and host proxy implemented directly on RDMA verbs, which allows the same kernels to run on any network backend (e.g. InfiniBand and on AWS EFA); we observe, surprisingly, that GPUDirect Async (IBGDA) yields little additional benefit over host-assisted GPU-initiated communication. We implement five kernels spanning tensor, sequence, and expert parallelism. On two 16-GPU H200 clusters, mKernel achieves speedups of up to 1.72x on GEMM+AllReduce and $1.88\times$ on Ring Attention.

A clinical agent benchmark can report the same verdict on identical inputs while the agent files a materially different order on each run. Such agents order tests, request medications and place referrals, yet benchmarks typically score one run per task and rarely ask whether identical inputs produce identical actions; MedAgentBench, the benchmark we use, scores a single attempt and says so. To measure this gap we introduce "same-input rerun", which replays a task with every input held fixed and compares the orders rather than the score, with six reliability metrics, and apply it to 1000 MedAgentBench runs across 50 tasks from its five write-capable families, two open-weight models below ten billion parameters quantised to four bits, and two temperatures. The study establishes that action-level divergence exists and can pass unrecorded by the score, not that any rate generalises. Under the 8B model at temperature 0.7, all 43 ordering groups emit a different set of orders across five identical runs, 26 emit the order on some runs and not others, and 28 record a different coded value, dose or analyte. In 22 of those 43 the benchmark reports the same failing verdict for materially different behaviour, as it does for all 10 divergent groups of the 4B model at 0.7. Orders also reach different endpoints across runs, one of which the record server rejects while the agent is told it succeeded. These findings motivate repeated-run evaluation, action-level stability reporting and execution-faithful environment feedback in clinical-agent benchmarks.

Large language models (LLMs) have demonstrated strong capabilities across a wide range of natural language processing tasks. However, conventional fine-tuning typically relies on centralized data collection, bringing in privacy concerns. Federated learning (FL) enables collaborative LLM fine-tuning without sharing raw client data, but its deployment over bandwidth-constrained wireless networks is hindered by the communication overhead of model-parameter transmission. Although Low-Rank Adaptation (LoRA) reduces the number of trainable parameters, its communication cost still increases with model scale. Knowledge distillation avoids parameter sharing via output logits, but token-level logits in LLMs incur high communication cost due to sequence length and vocabulary size. Reducing logits lowers the cost but weakens supervision and degrades accuracy. To address these limitations, we propose FLoKD, an adaptive knowledge-distillation framework for federated LoRA fine-tuning of LLMs over wireless networks, which communicates intermediate LoRA activations as the distillation signal rather than logits or full parameters. Since transmitting all blocks over the entire public dataset remains costly, we further propose a transformer block importance scoring framework that selectively transmits the most informative blocks, and two dataset selection strategies that discard public samples deviating from the local data distribution and prioritise those most informative for distillation. Extensive experiments across multiple generative language datasets, including WikiText-103, PTB, and Dialog, demonstrate that our proposed framework reduces communication overhead by 50-65% while achieving rapid convergence to competitive perplexity compared to baselines.

Safety research often focuses on model-generated harms, but users may also direct hostility, coercion, and adversarial pressure at models. Understanding how and when that occurs is essential for accurately interpreting model behaviour, alignment drift, and real-world deployment risks. In this paper, we audit 777K English LMSYS-Chat-1M conversations with two independent detectors: an eight-category lexicon for hostility directed at the model, and the dataset's moderation signal; and show that they capture different, weakly overlapping phenomena. The lexicon identifies insults, threats, and jailbreak coercion aimed at the assistant, while moderation flags are dominated by toxic-content solicitation rather than hostility at the model. Together, they mark about 5% of user turns; adjusting the narrower lexicon-harassment union for measured precision puts mistreatment aimed at the assistant at 0.90%. These absolute rates describe arena-style evaluation traffic and should not be read as deployment-wide base rates. We find that user hostility varies 13-fold across models, driven largely by who each model attracts rather than by model behaviour: first-turn hostility spreads far wider than post-response hostility, and more than fifteenfold separates the extremes even after deduplicating opening prompts. Within conversations, assistant apologies are consistently associated with higher odds of next-turn hostility under both detectors; the effect survives restricting to non-refused prior turns and to jailbreak-free conversations, and is positive in 20 of 23 models. Yet across models, more apologetic models receive less hostility overall. Finally, hostility also shows temporal structure, with coercive openings front-loading the first turn while affective hostility accumulates over a session. We release the lexicon, the detector cross-validation pipeline, and all derived tables.

Chemotherapy improves survival for some patients with breast cancer, but doctors cannot reliably predict who. Current guidelines rely on recurrence scores as a proxy for treatment benefit, which may contribute to the overprescription of chemotherapy. Here we present a causal multi-modal AI model that predicts personalized chemosensitivity using routinely collected pathology and clinical information. We developed our model on a multi-national dataset of 9,141 patients (twelve cohorts, nine countries) and evaluated it on another 1,994 patients (five cohorts, three countries). The model generated treatment-specific recurrence probabilities for each patient, with near-perfect calibration and strong prognostic discrimination across both 5- and 10-year follow-up horizons. Moreover, its chemotherapy benefit predictions demonstrated robust predictive performance, and out-performed existing recurrence-score-based tests. Compared to the standard of care, using the model to support personally tailored therapeutic decisions could reduce the number of patients receiving chemotherapy by 30% while achieving the same recurrence-free rate. Tumors predicted to be highly chemosensitive displayed concordant molecular and morphological programs of proliferation, cell cycle progression, and replication stress. The model's predictive capabilities transferred zero-shot to non-breast cancers, indicating our causal multi-modal AI approach may provide a universal strategy to predict treatment outcomes across cancer types.

Industrial maintenance systems involve multiple interacting assets and shared resources, making it challenging to balance reliability and operational cost using a single decision framework. While recent work has focused on reinforcement learning (RL) for maintenance scheduling, direct comparisons with planning approaches under identical settings remain limited. In this work, we empirically compare planning and RL for multi-asset bearing maintenance using run-to-failure data. We examine how these methods behave when balancing preventive maintenance against tolerable failures across a range of failure penalty scenarios. We observed a consistent behavioral difference driven by objective formulation. Planning enforces reliability as a hard constraint and produces zero-failure policies whose total cost is largely insensitive to the magnitude of failure penalties. RL agents optimize expected cost and often trade off preventive maintenance against occasional failures as penalties vary, resulting in lower costs under low-penalty regimes but persistent non-zero failures even when penalties are high. We also investigate lightweight constraint mechanisms, including reward shaping and action masking, to encourage RL's reliability. From a practical perspective, planning may be more suitable when strict reliability is required and deployment horizons are short, whereas RL may provide cost-efficient policies when limited failures are acceptable and long-run operational efficiency is prioritized. Overall, this study clarifies the trade-offs between reliability and cost in multi-asset maintenance and suggests that planning and RL are complementary approaches. Beyond these findings, the controlled benchmark protocol itself that unifies environment, cost model, and evaluation across paradigms, offers a reusable template for comparing decision-making approaches in other maintenance settings.

Efficient utilization of supply chain analytics for decision making remains a significant challenge for planners, as critical tasks such as database querying, key performance indicator (KPI) analysis, demand forecasting, and performance diagnosis require heterogeneous expertise spanning data engineering, operations research, and domain knowledge. In this work, we propose an agentic system for supply chain analytics that bridges the gap between business decision-making and technical expertise, where a coordinator agent interprets user intent and delegates sub-tasks to specialized agents. The system supports both exploratory analysis and deterministic workflows, enabling planners to transition between ad hoc questions and structured processes. Domain logic is encapsulated within specialist agents and prompts, yielding a scalable, modular, and auditable design and lowering the cost of functional extension through prompt-centric development. We evaluate the proposed architecture on a test environment that replicates multi-echelon inventory management operations. Results show that our multi-agent design achieves a 90% accuracy, which is competitive with a single agent baseline while reducing input token usage by roughly fourfold, substantially improving scalability and cost-efficiency. Furthermore, we provide case studies to demonstrate interpretable suboptimality detection and automated forecast optimization, illustrating how agentic architectures can effectively combine open-ended exploratory analysis and deterministic supply chain analytics workflows, and provide a practical pathway toward more accessible and extensible decision-support systems.

Large Language Models (LLMs) with function-calling capabilities are becoming critical for modern agentic AI systems. Nevertheless, current deployments typically route inferences to powerful cloud-based models, incurring significant energy use and carbon emissions. We address this sustainability challenge with a carbon-aware routing framework that distributes function-calling queries across a three-tier edge-cloud architecture, combining edge and cloud LLMs on heterogeneous hardware. At its core, a lightweight k-NN predictor operating in a unified semantic-lexical embedding space estimates query-specific accuracy, delay, and power consumption on each edge tier. These predictions are then combined with real-time grid carbon intensity to route every query to the lowest-emission tier capable of executing it successfully. Evaluated on state-of-the-art function-calling benchmarks and LLM families, our framework matches cloud-level accuracy while reducing operational carbon emissions by $4\times$ on average.