每天知道
你的领域
出了什么

Large Language Models (LLMs) have demonstrated strong capabilities in code generation, but their outputs frequently contain syntax or type errors that result in compilation failures. Constrained decoding has been proposed as a solution to mitigate compilation errors by construction, improving functional correctness as a byproduct. However, previous works overlook a critical aspect of constrained decoding: the alignment between constrainer (e.g., types), language model and the target specification language (e.g., TypeScript). Misalignment is caused by the constrainer being incomplete--rejecting programs that belong to the target--or unsound--allowing programs that are not part of the target. The bias created by incompleteness distorts the language model distribution, and can be detrimental for code generation. We evaluate this hypothesis using seven language models, two target languages, two constrainers, enforcing types and syntax during decoding, and we study how language models react to varying levels of incompleteness. On three benchmarks, when the constrainer is incomplete, unconstrained decoding significantly outperforms constrained decoding in terms of functional correctness. Incompleteness pushes the model into low-probability regions of the program space, causing the generation to frequently time out, and reducing functional correctness by up to 97%. These contributions make the community aware of the negative effects of misalignment in constrained decoding, and provide quantitative insights on how to design constrainers that are beneficial for code generation systems with formal guarantees.

Large language models (LLMs) store and recall factual knowledge, yet the precise mechanism of how entity representations are transformed to enable specific attribute retrieval remains underexplored. In this work, we investigate this mechanism through the lens of an "attribute-computation path"-a sequence of computational steps over the entity representation required to elicit a target attribute. We then propose an iterative patching protocol to identify a minimal subset of layers necessary for this computation. Applying our method to LLaMA 3.1 8B and Qwen3 8B, we find that these paths are non-contiguous, often skipping layers, and that models possess multiple, functionally-equivalent paths for the same entity and fact, highlighting a high degree of redundancy in attribute computation. This implies that knowledge computation is highly distributed, potentially explaining the localization-editing mismatch and suggesting that knowledge storage and retrieval in LLMs is far from being well understood.

We introduce RocketPFN, a training-free pipeline for time series classification that combines random convolutional feature extraction (Rocket) with in-context classification via a pretrained tabular foundation model (TabPFN v2.5). On 92 UCR datasets (30-resample protocol), RocketPFN matches HC2, the strongest published method on the archive, in mean accuracy (both 0.900, Wilcoxon p=0.50), with no training on the target data and a median inference time of 30 seconds per fold. It also significantly outperforms every individual classifier in the HC2 ensemble. On UEA (20 datasets) the difference is likewise not statistically significant. A separate comparison concerns TSC foundation models: when paired with the same downstream classifier, MOMENT, Mantis, and MantisV2 are all significantly outperformed by RocketPFN using fewer extracted features and no learned parameters (p<0.001 in each case). This holds even when the encoders were pretrained on corpora that include the UCR training samples. We propose this two-stage pipeline as a reference point for evaluating zero-shot TSC foundation models.

Variable-order Markov models preserve local symbolic syntax by adapting context length, but long continuations can enter recurring high-order "tunnels": repeated suffixes, locally periodic passages, or copied fragments longer than the formal Markov order. This paper introduces signed pattern control for variable-order Markov generation with BP-Regular sampling. A weighted recurrence automaton computes an activation R for a chosen family of target patterns, and belief propagation samples exactly from P_beta(x) proportional to P_0(x) exp(beta R(x)). Negative coupling makes the target patterns costly during sampling; positive coupling rewards the same patterns and turns them into controlled attractors. The target family may be mined online from overactive generated material, supplied by a score or style vocabulary, or designed as an experimental probe. The main experiments use the online homeostatic case, choosing patterns that become overactive in the sampling history. On six duration-bearing monophonic sources, including Bach and Telemann material, the negative branch reduces generated 8-gram self-reuse, increases the effective number of generated 8-grams, and increases coverage of training-supported 4-gram contexts while preserving substantial lower-order support. A pitch-sequence replication on five Weimar Jazz Database solos gives the same anti-reuse signature outside Baroque material. The same signed mechanism also provides a positive branch for probing attractor basins, phase transitions, and hysteresis in the underlying variable-order model.

A finite agent, a machine's digital twin, or any bounded reasoner, sees a fixed, noisy world only through finite sensors, so its coherent output is not a point but a belief: a probability density over states (the Bayes posterior). Certainty is denied twice, by observation (Cramer-Rao) and by physics (Landauer), both diverging at the boundary where the Fisher information blows up. We turn this finiteness into geometry: belief-cost geometry, the geometry of what it costs to change one's mind. The cost metric is optimal transport in Wasserstein space, conformally reweighted by Fisher information (the price of the precision at stake), $\tilde g_{e,U}=2(e+U)\,g_{W_2}$ with relief $U$. It rests on two posed postulates: that revision cost is a scalar price on transport (the arena), and that the price is honest, one nat costs the same length everywhere (eikonal). Honesty selects the Fisher reweighting because transport demotes the Fisher information from the metric ruler of distinguishability (its role in Fisher-Rao information geometry) to the slope of entropy. Three results follow on the conformal class: a wall, a well-posed inference pushes certainty to infinite cost-distance once the relief dominates the Fisher information (necessity conjectured beyond power laws); an honest family, the eikonal price is equivalent to $U=cJ$, the Fisher family; and a rigidity (essentially location-scale), these geometries are hyperbolic and the Stam bound crowns the Gaussian as the most hyperbolic location-scale belief (ranking at $e=0$), the value $-1/4$ being one image of a relativity of cost. The cost of reaching a given precision then has a geometric floor diverging at certainty. Thermodynamics fixes the cost unit (one nat costs $k_BT$ at the wall) and motivates the framework; the results are geometric, in nats.

Mixture-of-Experts (MoE) language models are often described as ideal for resource-constrained inference. Each token activates only a small subset of experts, so the per-token compute cost, in floating-point operations (FLOPs), resembles that of a much smaller dense model. Whether that FLOP advantage survives in practice is far less clear. We ask whether MoE models actually run faster and cheaper than comparable dense models on consumer-grade and edge hardware. We benchmark OLMoE-1B-7B (1.3 B active of 6.9 B total) against three dense baselines on an Apple M2 Pro and an NVIDIA Jetson Orin Nano 8 GB through llama.cpp, measuring throughput, memory, and on-device energy. The answer is device-dependent: OLMoE's active-parameter advantage is only partly realised on the laptop (~10% behind the same-active Llama-3.2-1B) and erodes on the edge device (~31% behind, at 2.1$\times$ the energy per token, with peak memory at the 8 GB ceiling). Patching llama.cpp to time the decode graph node-by-node shows routing accounts for under 9% of MoE-block compute on the cleaner edge backend, so the gap reflects total-parameter memory footprint, expert dispatch, and KV-cache pressure rather than routing. The implication is that on bandwidth-bound edge hardware, inference cost tracks total parameters, not active ones, and sparse activation does not buy back what the device is constrained on. These findings are bounded to one MoE model at this parameter scale and two devices, and we release the full measurement harness and per-run data.

The Helmholtz equation governs wave propagation in acoustics, electromagnetics, and seismology, but its indefinite nature makes it difficult to solve with iterative methods. Domain decomposition methods are a natural fit for massively parallel architectures, yet mapping efficient Helmholtz solvers onto modern GPUs remains a challenge. We address both with two key contributions: (1) a block-level domain decomposition scheme, in which each subdomain is assigned to a single thread block and all solves run concurrently in a single kernel launch, and (2) WaveHoltz as the subdomain solver. WaveHoltz is a fixed-point iteration that is uniquely well-suited to the GPU execution model due to its minimal memory footprint and no reduction operations. Together, these eliminate device-level synchronizations and replace global memory traffic with shared memory and register-level operations, keeping subdomain data largely resident in L1 and L2 cache. We explore two threading strategies: one degree of freedom per thread for small subdomains, and multiple degrees of freedom per thread for larger ones. Benchmarks of our CUDA based implementation on a NVIDIA A100 show that WaveHoltz achieves 2x-25x speedup over MINRES, with the advantage growing with subdomain size. Crucially, evaluating the subdomain solver in single rather than double precision yields an additional 2x-10x speedup--a benefit largely unattainable by MINRES due to loss of Krylov vector orthogonality under reduced precision.

We present R2HandoverSim, a simulation benchmark for robot-to-human (R2H) object handovers. Although R2H handover methods have advanced rapidly, the lack of standardized evaluation protocols impedes objective comparison. Our benchmark enables reproducible evaluation by systematically comparing four baselines on their predicted shared grasp poses. We conduct a user study with 30 participants, analyze baseline performance, and show that simulation results correlate with real-world evaluation outcomes. Crucially, five complementary metrics (planning feasibility, reachability, grasp stability, grasp affordance, and safety) better reflect user-perceived handover quality than overall success rate alone. Website and code: https://robot-future.github.io/r2handoversim/.

As Large Language Models (LLMs) deploy into mission-critical domains (e.g., finance, medicine, and law), output reproducibility has become a strict system requirement. While practitioners use greedy decoding to eliminate algorithmic stochasticity, empirical deployments with 16-bit precisions still exhibit catastrophic output divergence across heterogeneous GPUs. Through SASS-level profiling, we reveal that this inconsistency is fundamentally driven by truncation errors introduced during downcasting at kernel boundaries. However, achieving reproducibility via a global FP32 pipeline incurs prohibitive system penalties: bypassing 16-bit hardware accelerators hurts compute efficiency, while upcasting the KV cache doubles memory overhead. To bridge this gap, we propose Hybrid Error ALleviation (HEAL), a targeted intervention that approximates FP32 precision while resolving hardware constraints through two targeted mechanisms. First, recognizing that floating-point formats underutilize their bit-width for Q, K, V tensors, HEAL applies INT16 quantization that preserves numerical stability without expanding the KV cache footprint. Second, HEAL synthesizes high-precision matrix multiplications via an algebraic error compensation strategy, executing entirely on high-throughput 16-bit Tensor Cores. To evaluate our approach practically, we introduce MCR-Bench, a benchmark targeting reproducibility in mission-critical tasks. HEAL achieves the same level of reproducibility on downstream tasks as the FP32 baseline while reducing the performance overhead by up to 7.1x.

Psycholinguistics studies show that human readers fall for coherence illusions: an incoherent discourse can seem coherent simply because a distractor matches what comes next. We investigate whether Dutch language models (6 monolingual and 4 multilingual) show the same behavior on texts that link back to earlier context with words such as 'again' and 'too'. First, we find that surprisal at the critical word tracks human acceptability judgments and eye-tracking data. Models are more surprised by incoherent continuations, but a matching distractor in the prior context reduces this surprisal. Second, attention entropy at the critical position identifies heads that behave differently under coherence vs. incoherence. We find that ablating these heads shows transfer effects across experiments, suggesting a shared mechanism. Third, we introduce energy from the associative-memory literature as a metric to quantify discourse coherence. Taken together, our results show that coherence illusions arise in Dutch LLMs, with entropy and energy exposing mechanisms that operate across settings.

Dysarthria severity assessment is essential for therapy planning and longitudinal monitoring, yet manual perceptual rating is time-consuming and variable across clinicians. Although deep learning models achieve strong performance, their black-box nature limits clinical adoption. Existing speech explainability methods typically provide acoustic feature importance scores that are difficult for end-users to interpret. We propose an influence-based, instance-level explainability framework that explains each decision through supportive and competing training samples. Using gradient-based influence approximations, we compute per-utterance influence scores to identify supportive and competing training samples for each prediction. Controlled deletion experiments from 5 to 20 percent validate the explanations, showing that removing highly influential samples systematically shifts predictions. This approach provides auditable explanations by linking decisions to perceptible reference cases.

Modern deep neural network architectures are trained via backpropagation, which requires errors to be sequentially propagated through all layers before parameters can be updated. This introduces two limitations: locking, where layer-wise updates are strictly interdependent and cannot proceed in parallel, and the weight transport problem, which requires symmetric forward and backward pathways for exact gradient computation. These constraints restrict parallelism, increase memory and communication overhead, and pose challenges for scalable learning. In this work, we propose Hierarchical Block-Local Learning (HBLL), a framework that decomposes deep neural networks into hierarchically linked blocks trained using local learning objectives derived from variational principles, eliminating the need for full end-to-end backpropagation while maintaining effective information propagation across the network. HBLL is the first algorithm that is able to train deep neural networks in $\mathcal{O}(\log N)$ parallel time complexity, where $N$ is the number of network layers. We show that HBLL implicitly defines a family of subnetworks corresponding to different hierarchical paths, enabling flexible inference with different effective numbers of layers. We evaluate HBLL on a set of challenging vision and language modeling tasks, achieving competitive performance. We also extend HBLL to recurrent sequence architectures, applying to settings that otherwise rely on backpropagation through time.

The open-source ecosystem on GitHub lacks a systematic hierarchical taxonomy of software repositories. GitHub Topics, the dominant organizational mechanism, is flat, inconsistent, and covers only 67% of projects. We present ATLAS, the first framework that automatically constructs a hierarchical taxonomy for software repositories and classifies projects into it end-to-end. By combining LLM global knowledge with real repository distributions, ATLAS proposes meaningful splitting dimensions and iteratively corrects those that fail to accommodate real projects. A Designer Agent proposes splitting dimensions while a Classifier Agent assigns repositories; a self-corrective refinement loop uses classification failures to drive dimension revision through escalating strategies. We evaluate ATLAS on 54,387 GitHub repositories against six baselines spanning four paradigms, two downstream tasks, and three model families. On a stratified 2,001-repository benchmark, ATLAS achieves a Taxonomy Quality F-score (TQF) of 83.13%, outperforming the best baseline by 15 percentage points (on the full 54k corpus the approximate TQF is 73.0%, a gap driven by Path Granularity's all-or-nothing scoring on longer paths rather than lower classification accuracy). It is the only method to simultaneously achieve high structural quality and high practical applicability. On downstream tasks, ATLAS enables alternative discovery with P@1 = 85.71%, surpassing even human-curated lists (62.34%), and achieves the highest P@1 for repository retrieval. The taxonomy further reveals structural ecosystem trends that are difficult to obtain from flat tags or similarity methods: the shift from libraries to AI/ML applications (now 61% of newly community-adopted projects) becomes visible only through hierarchical, type-based categorization. An interactive taxonomy explorer is available at https://atlas-taxonomy.netlify.app/

The evaluation of vision-language models (VLMs) for chest X-ray (CXR) analysis has largely been limited to disease-presence classification without visual grounding. Such evaluations fail to verify the expert-level lesion perception necessary to ensure the clinical reliability of VLMs. To address these limitations, we introduce CheXpercept, a sequential, multi-level perception benchmark that mirrors a radiologist's cognitive workflow across coarse-level detection, fine-level contour evaluation and revision, and semantic-level attribute extraction. To ensure high clinical fidelity at scale, we construct the dataset using a semi-automated generation pipeline paired with a review by six medical experts. CheXpercept contains 10,400 QA items derived from 2,100 CXRs, covering seven clinically critical pulmonary and cardiac lesions. To demonstrate the current landscape of VLM perception, we benchmark 14 general and medical VLMs on CheXpercept. The models achieve adequate performance only at the coarse level, with accuracy degrading precipitously on deeper visual tasks. Notably, medical VLMs show almost no perceptual advantage over their general-domain counterparts, highlighting a systemic flaw in current domain adaptation. The code and dataset will be publicly available.

The empirical foundation of cyber deception relies on human-centered hypotheses, but the rapid emergence of autonomous, AI-enabled attackers challenges whether this foundation transfers to AI agents. To address this, we introduce an automated evaluation framework adapted from the Honeyquest instrument to assess LLM attacker judgment at scale. Our 21-LLM cohort spanned 10 providers, diverse architectures and specializations, open- and closed-weight models, and parameter scales from 8B to over 1T. We evaluated the performance of this LLM cohort (yielding 10,962 responses) against the 47-participant human baseline across an identical set of 174 reconnaissance queries. Our empirical evaluation reveals three key findings that establish LLMs as a distinct attacker class: (1) every model in our cohort falls for deceptive traps at a significantly higher rate than human attackers; (2) the defensive attention-diversion effect observed in humans is statistically absent in our LLM cohort; and (3) a critical recognition-action gap, where LLMs successfully articulate trap recognition in their reasoning but exploit the deceptive elements anyway 73.4% of the time. Across the 21 models, trap recognition in reasoning text did not predict fell-for-trap behavior (Spearman $r = +0.08$, $p = 0.73$). Ultimately, these findings demonstrate that human-centered deception hypotheses do not reliably transfer to AI attackers, highlighting the critical need for new research into AI-native active defense frameworks.

Procedural mistake detection is important for quality control and user assistance across many disciplines. Recent work in this field has achieved significant gains by using the reasoning capabilities of Video-Language Models (VLMs) as components within multi-stage pipelines, which consist of separate modules for supervised temporal action segmentation, error detection, and explainability. Consequently, they remain dependent on tailored training datasets and require task-specific training, limiting their wider applicability. To remedy this, we introduce zero-shot procedural mistake detection and propose a unified Zero-shot Procedural Mistake detection (ZeProM) framework that jointly solves procedural mistake detection and temporal action segmentation with a single pre-trained VLM. By evaluating our framework on two canonical mistake detection benchmarks, EgoPER and CaptainCook4D, we find that ZeProM can perform these tasks successfully, while approaching, or even outperforming, the performance of fully supervised methods. For instance, we achieve a 4.4 point improvement in EDA and a 2.0 point improvement in F1@.5 on average over all five EgoPER tasks compared to the strongest supervised methods. Overall, our results show the potential of unified methods for procedural mistake detection, and we hope this will steer the field away from highly complex pipelines and toward more generally applicable solutions.

Code review requires reviewers to understand the core intent of code changes, which becomes difficult when a commit modifies multiple classes. In such commits, one or more primarily modified classes, referred to as salient classes, may induce modifications in other classes. Accurate identification of salient classes offers reviewers an effective entry point to navigate code changes and facilitates program comprehension. Existing state-of-the-art approaches rely on complex program-analysis procedures, including Abstract Syntax Tree (AST) parsing, class relation extraction, handcrafted feature engineering, or dependency graph construction. To this end, we study whether language models (LMs) can identify salient classes directly from commits without feature engineering, graph construction, or training. We first construct a new dataset ApacheJavaCM, derived from the ApacheCM dataset, containing 7,911 commits and 25,914 labeled classes. On this dataset, we systematically evaluate whether LMs can identify salient classes directly from commits and compare with the strongest reproducible state-of-the-art (SOTA) baseline. The evaluation covers two large language models (LLMs), GPT-5.4 and DeepSeek-V3.2, one small language model (SLM), Qwen3.5-9B, and three prompting strategies: zero-shot, few-shot, and chain-of-thought. The LMs substantially outperform the baseline while remaining stable across commit characteristics and selected LMs. We also found that, for salient class identification tasks, a 9B-parameter open-source SLM, Qwen3.5-9B, under few-shot prompting, achieves performance comparable to that of a much larger closed-source LLM, GPT-5.4. These results suggest that lightweight, locally deployable SLMs are sufficient for industrial salient class identification tasks and can reduce both cost and privacy barriers associated with relying on closed-source LLMs.

Robotic inspection relies on accurate sensing to acquire high-fidelity geometric measurements for defect detection and metrology. While prior work has focused on robot motion and viewpoint planning, how to configure sensing parameters remains largely underexplored, despite their decisive impact on measurement quality. We propose SenseHD, a robotic sensing system that formulates scanner configuration as an instruction-conditioned sensing decision. Instead of predicting precise parameter values, SenseHD treats sensing parameters as discrete sensing actions and selects stable sensing regimes through hyperdimensional associative memory. Experiments on a real robotic inspection platform demonstrate that SenseHD robustly selects appropriate sensing configurations and significantly improves inspection reliability, while remaining lightweight and efficient compared to baseline methods.

Large language models can produce confident but protocol-invalid answers in domains where procedural compliance is critical. This paper presents Answer Engineering, a deterministic runtime and authoring layer that applies localized rule-guided interventions to the visible reasoning trajectory during standard autoregressive generation, without retraining, modifying model weights, or performing global search. The method is evaluated on a controlled clinical benchmark for sudden sensorineural hearing loss (SSNHL), where correct management depends on protocol-consistent interpretation of symptom timing, Weber/Rinne tuning-fork findings, and otoscopic findings. In the benchmark, step-by-step reasoning shifted rather than eliminated errors: compliant outcomes for SSNHL decreased from 54.5% under unguided generation to 25.1%, while acceptance on the conductive contrast condition increased from 1.6% to 58.9%. Local trajectory editing increased SSNHL compliance to 83.5% and conductive-case adherence to 77.9%, raising balanced accuracy from 42.0% under reasoning-only generation to 80.7%. The results support a systems-level view in which protocol adherence can be improved through auditable runtime control of reasoning trajectories, while also identifying limitations caused by rule coverage, trigger reliability, and persistent diagnosis-first generation dynamics.

Conversational AI agents in Customer Experience (CX) typically suffer from a Closed-World Constraint, ignoring high-velocity external shifts like viral trends or outages. Ad-hoc web search attempts to bridge this gap but often introduce prohibitive latency and context poisoning. We introduce PulseCX, a framework that decouples knowledge acquisition from consumption. Adopting a structure-first paradigm, PulseCX employs an asynchronous agent to linearize signals into a Decay-Aware Temporal Knowledge Graph (DA-TKG) governed by reinforcement--decay dynamics to actively manage information lifecycles. By coupling this self-evolving memory with hierarchical intent gating, PulseCX removes synchronous search bottlenecks (<10ms overhead) and drives significant gains in Intent Resolution (IRR) and Customer Satisfaction (s-CSAT) in dynamic environments.

LLM agents increasingly solve local tasks through command-line and CLI-based harness interfaces, including code editing, repository inspection, data analysis, and file workflows. Existing evaluations often emphasize task success, but deployed local agents are not models alone: the CLI mediates prompts, context replay, tool outputs, file access, terminal observations, and stopping behavior. As a result, the same model can produce different success, token, and cost profiles under different CLIs. We introduce AGENTMETER, a benchmark for evaluating model-CLI matching in CLI-mediated local task-solving agents, together with AgentMeter Score (AMS), a success-anchored, cost-aware metric over calibrated task-effort tiers. AgentMeter uses Benchmark90 as the full validation set and Core30 as a lower-cost subset for expanded comparison across 24 complete model-CLI configurations. On Core30, common deployment criteria select different configurations: highest Pass/30 selects GLM-5.1 with qwen-coder, lowest Tok./Pass selects GPT-5.3-Codex with kimi-cli, lowest billable USD/Pass selects Qwen3.6+ with Codex, while highest AMS selects Qwen3.6+ with kimi-cli. Benchmark90 validation preserves the Top-1 configuration and Top-3 set, with Spearman correlation 0.765, Kendall correlation 0.567, and AMS MAE 0.0383. These results show that model choice and CLI choice should not be decoupled, and that model-CLI configurations should be evaluated as the deployed unit.

Automated test generation for telecom software systems and networks has advanced significantly with the adoption of machine learning and rule-based approaches. However, most existing solutions generate static test suites against a snapshot of the system; as code, configurations, topologies, and key performance indicators (KPIs) evolve, these tests quickly become outdated or misaligned with the live system. There is currently no widely adopted solution that continuously detects fine-grained changes and selectively adapts only the affected tests without regenerating entire test suites. This paper presents a context-aware generative AI framework for automated telecom test script generation that treats testing as a continuously adapting process driven by the current state of the system rather than a static artifact. The central contribution is delta-conditioned test generation over a live knowledge graph: our approach employs a continuously updated knowledge graph (KG) as a single source of truth, a delta engine for fine-grained change detection, and a KG-guided generative AI agent, operating via the Model Context Protocol (MCP), to create, update, or retire test cases automatically. We further integrate Retrieval-Augmented Generation (RAG) to enrich reasoning with telecom-domain knowledge and historical artifacts. We demonstrate applicability across software-system and telecom-network use cases, including a Python-based KPI monitoring application managed in GitLab, and show how the framework reduces manual effort, improves test relevance, and accelerates test cycles.

This paper evaluates the performance of twelve Small Language Models (SLMs) on Arabic natural language processing tasks. The study introduces a benchmark of 240 Arabic test items distributed across eight domains and ten language skills, covering both comprehension-oriented and generation-oriented tasks. All models were evaluated under a controlled zero-shot setting using a standardized Arabic-only prompt template. Model responses were assessed through a multi-model LLM-as-a-judge framework involving GPT-4.1 Mini, Claude Haiku 4.5, and DeepSeek-Chat, with scores aggregated across judges and analyzed by task, skill, and model family. The results show that Gemma 3 (12B) achieved the highest overall score (4.548/5), followed by Aya and C4AI Command Arabic. The observed results suggest that model size alone does not explain Arabic SLM performance. Models with stronger Arabic alignment and more reliable instruction-following behavior tended to perform better across tasks. Common failure patterns among lower-performing models include prompt leakage, hallucination, language drift, incomplete generation, and weak task adherence. Overall, the benchmark provides a structured reference for evaluating compact Arabic language models and supports future work on efficient, reliable, and culturally appropriate Arabic AI systems.

Agentic retrieval-augmented generation (RAG) systems combine iterative reasoning loops, query decomposition, and adaptive retrieval to tackle multi-hop question answering. However, the contribution of each component remains poorly understood, particularly under resource-constrained settings using only local language models. Many agentic designs add adaptive retrieval routing and deeper retrieval loops on the assumption that the added complexity helps. To test whether it does, we run a controlled ablation study of a full agentic RAG pipeline evaluated on 5,000 questions from the HotpotQA distractor development set using a local 7B parameter model (Qwen2.5-7B-Instruct). Our full pipeline achieves EM=53.2% and F1=61.6%, compared to a single-pass dense-retrieval baseline of EM=43.1% and F1=54.0%. Across eight ablation conditions, we find that: (1) fixed hybrid retrieval via reciprocal rank fusion consistently outperforms rule-based adaptive routing (+1.8 EM, +1.9 F1), as the routing heuristic over-routes to BM25 by firing on named entities present in nearly all multi-hop sub-questions; (2) two retrieval iterations over the decomposed sub-questions capture 95% of the gains of five, with no meaningful benefit from deeper loops; and (3) query decomposition and cross-encoder reranking each contribute statistically significant but smaller gains (p<0.01 and p<0.001 respectively). Taken together, on a fixed local-model budget, the simpler and fixed choices turn out to be competitive with or better than their adaptive versions: most of the gain comes from running a short retrieval loop, not from adaptive routing or from many iterations. We use no proprietary APIs or large-scale compute.

Agentic AI systems orchestrate multiple LLM-based agents through workflow architectures that coordinate decisions, tools, and external actions. While current platforms emphasize runtime safeguards, little support exists for verifying workflows during system design. From a Modeling & Simulation perspective, this gap is analogous to composing conceptual models without verifying whether their building blocks interact coherently. We propose a design-time verification approach that models agentic workflows as compositions of reusable building blocks and checks their compatibility through twelve structural rules. We implemented these rules in a software prototype and evaluated them using two openly released datasets: 48 workflows with known design flaws and 168 variants that preserve workflow logic but alter graph structure. Results show that our verifier reliably detects violations even when flawed designs are obscured through structural transformations such as splitting tasks between agents. Future works could combine our verification with community repositories of building blocks to compose safe agentic workflows.

Large language models (LLMs) have been proposed as hyperparameter-optimization (HPO) advisors that "warm-start" search from prior knowledge, proposing strong configurations in very few evaluations. We test that claim under a budget-matched, multi-seed protocol on eight PMLB tabular benchmarks, comparing an LLM advisor (LLM-OptFlow) against four classical baselines (random search, Optuna-TPE, Gaussian-process Bayesian optimization, and successive halving) over one shared search space, with paired tests and bootstrap 95% CIs across 8 x 5 = 40 (task, seed) units. The finding is cautionary. The advisor's strong first point is not an LLM output at all: like prior LLM-HPO systems the loop is seeded with a fixed default configuration, evaluated before any model call, which alone reaches 88.7% mean best-CV, identical to within 0.01 pp across all seven advisor models tested. The LLM's own proposals add only +0.40 pp of cross-validation accuracy over that seed and nothing on held-out test (LLM-Default = -0.01 pp, p = 0.92). When the same seed is granted to classical search, the apparent lead collapses: against seeded random search it leads by +0.20 pp at 2 evaluations, is tied by 5, and is behind by 12 (-0.37 pp). Without the seed, classical search ties the advisor by 12 evaluations and beats it by 40 (+0.6 to +0.8 pp, p <= 1e-4). Two LLM-specific behaviors survive: a single-task exploration failure (vehicle), and a rule-based confidence filter that removes ~33% of wasted compute without changing accuracy. The recommendation is deflationary: on tabular HPO, seed classical search with a sensible default; an LLM advisor adds no measurable generalization benefit and is overtaken within a handful of evaluations. We release the harness and a script that reproduces every statistic.

We present OmniV2X, a generative foundation model for vehicle-to-everything (V2X) cooperative driving. The model directly interprets independent context sequences comprising multi-modal and multi-agent observations. The new design mitigates the computational cost of dense 3D perception, the vulnerability to data scarcity in cooperative scenarios, and the poor compliance with standardized messaging in existing methods that fuse multi-modal inputs into a shared representation. For training, we present an end-to-end supervised pipeline using a downstream trajectory generation loss, in which a high-capacity generative sequence planner implicitly learns to steer the model and leverage multi-modal inputs via cross-attention injection. As a foundation model, we demonstrate that OmniV2X pre-trained on large-scale single-agent planning datasets can efficiently adapt to cooperative environments by integrating the conditioning context with lightweight, standard-compliant V2X tokens. Evaluated on the DAIR-V2X-Seq dataset, OmniV2X outperforms existing end-to-end cooperative driving baselines, achieving state-of-the-art performance with less than 10% of the fine-tune V2X dataset and less than 1% of the communication bandwidth. We conduct comprehensive evaluations to demonstrate its computational efficiency and robustness under real-world constraints.

Transformers are AI's workhorse with strong performance in modeling sequential data, but their computational cost becomes prohibitive when processing long sequences. We target long-horizon streaming vision and robotics applications like map-free pose estimation, where it is particularly impractical to store and maintain a history of observations. Recurrent Transformers address this limitation by maintaining fixed-size memory but their performance lags behind that of transformers operating over the full observation history. We argue that this gap does not stem from architectural limitations, but from differences in how these models learn to compress past information. Without access to an observation history, recurrent models must explicitly decide what to retain in memory at each step, a significantly harder learning problem. In this work, we propose a distillation approach that transfers the compression strategy of a classical full-history transformer to a recurrent variant. We enable this by designing a teacher model that explicitly compresses its observation history into a fixed-size bottleneck representation. By directly supervising the student's memory with this bottleneck representation, we align the two compression mechanisms. We show that this approach allows to train a recurrent latent robotic memory with linear-time complexity while substantially narrowing the performance gap to full-history transformers.

Maintainability is a core dimension of software engineering, shaping how code is written, reviewed, and developed over time. While coding agents have demonstrated strong performance on single-issue tasks, it remains unclear how maintainable their code is when future agents build on top of it, potentially leading to compounding downstream effects. We investigate how agent code compares to human code in these maintenance settings, presenting CodeThread, a framework to construct controlled experiments from repository-level coding benchmarks. Applying CodeThread to four frontier coding agents and four benchmarks, we find that agents are less effective at resolving tasks when building on agent code compared to human code, with task resolve rate drops of up to 13.1%. Regression analysis reveals that many traditional software engineering maintainability metrics do not explain this difference. Instead, the clearest signals are subtler behavioral differences in agent code, such as changes to input validation and error handling, along with differences in downstream code size and task difficulty. These findings highlight the need to evaluate these systems not only by immediate task resolution but also by code maintainability, and point to potential sources of downstream errors introduced by agent code.

Vision-Language Models (VLMs) for embodied navigation rely on selecting a fixed number of frames from a growing trajectory history. As episodes extend, this selection grows increasingly sparse, yet prior work shows no accuracy gain when scaling from 8 to 64 frames, suggesting the bottleneck is not frame quantity but the representation itself. Sparse frame selection cannot capture the structured behavioral signal that long-horizon reasoning requires: turning patterns, cumulative displacement, and path topology. We introduce BIT-Nav (Brain-Inspired Trajectory Memory for Navigation), a framework that augments frozen VLM navigation pipelines with a compact learned trajectory memory. Motivated by hippocampal path integration, where spatial experience is compressed into structured episodic traces rather than stored as raw sensory replay, BIT-Nav trains a Bi-GRU encoder over action and relative pose sequences via a multi-positive InfoNCE contrastive objective on trajectory prefixes sharing the same behavioral intent. The resulting embedding is projected into the VLM token space via a lightweight MLP and injected as a single memory token at each decision step, conditioning the model on structured motion history at constant token cost regardless of episode length