Search research

This paper presents a self-supervised representation learning framework for understanding 3D skeleton-based human motion in soccer, using future motion prediction as the learning objective. Since human motion is inherently uncertain, accounting for multiple plausible futures is essential for capturing the underlying motion dynamics and learning effective representations. To this end, we introduce a conditioning module for motion prediction that models a probabilistic distribution over discretized future motions in 3D Euclidean space, learning multimodality with explicit supervision from future trajectories. Experiments on large-scale soccer player tracking data show that our approach substantially improves motion prediction accuracy. Moreover, the learned representations effectively transfer to multiple soccer downstream applications, demonstrating strong cross-task generalization.

Shani et al. (2026) show that LLM representations broadly recover human category boundaries, while failing to reflect fine-grained typicality structure. Their analysis uses cosine similarity over dense model representations. We revisit their approach using overlap over active sparse autoencoder (SAE) latent sets as a more interpretable similarity measure. We first verify that this set-level measure is meaningful: SAE latent sets can recover union-like compositional structure in controlled toy models and induce semantically coherent neighborhoods in natural text. Extending the human-concepts analysis to SAE set similarities, we find that SAE activation sets do not recover human category boundaries or within-category typicality more faithfully than dense embeddings or residual-stream states, but instead track model-internal similarity structure. To probe this gap further, we study active latent sets under well-controlled semantic modifications, revealing a substantial mismatch between human judgements of conceptual change and change in the SAE active set. We interpret this as evidence that, outside idealised settings, SAE features do not compose via simple bag-of-features semantics.

We propose that a model's uncertainty about a token is reflected not only in the breadth of its output distribution but also in whether a confident prediction is \emph{fragile} under perturbation of its attention pathways. We instantiate this as ASMI (Attention-Subnetwork Mutual Information), a training-free estimator that masks attention heads and measures the BALD mutual information among the resulting subnetworks, with a semantic-agreement kernel to discount surface-form disagreement. The signal is not a restatement of output confidence: on grounded QA an out-of-fold test shows it adds error-predictive information beyond single-pass confidence and entropy, concentrated in \emph{confident-but-fragile} predictions, where acting on it roughly halves the retained error of a confidence filter. The distinctness is regime-graded, so ASMI predicts its own domain of applicability, strong where answers are routed through provided context and bounded by design where they are recalled from parametric knowledge. Sem-ASMI reads the signal from a single greedy response, without the stochastic generations the strongest baselines require, and ties or beats Semantic Entropy on ten of the twelve grounded benchmark-backbone settings. Across the same twelve settings, the best ASMI variant, typically the adaptive one reusing the ten samples already drawn for the baselines, ties or leads the strongest baseline in eight, significantly in three under a paired test. On parametric QA all variants revert to or below the zero-cost MSP baseline, exactly as predicted, and the estimates are near-deterministic across reruns. A head-level analysis shows that what tracks this boundary is not the presence of head-level fragility but whether that fragility couples to errors.

Assessing catheter and tube placement on chest X-rays is safety-critical yet tedious and error-prone. Current deep learning methods either classify placement globally -- losing track of which device is where -- or segment all devices into a single mask, making per-device assessment impossible when catheters overlap. We introduce UCompCXR, a compositional framework that detects local catheter fragments, associates them into device instances via graph-based clustering, fuses per-fragment tip predictions through precision-weighted Gaussian estimation, and classifies placement per device. On the RANZCR CLiP dataset (30,083 images, 5-fold patient-level CV with bootstrap CIs), UCompCXR detects 26% more devices than a strong multi-task baseline sharing the same MobileNetV3 backbone, with 75% fewer false positives and well-calibrated tip uncertainty (95% coverage = 0.948). The aggregate tip error rises -- but only because the model finds devices the baseline misses entirely, especially nasogastric tubes. On matched devices, catastrophic localization failures drop substantially. At 2.27M parameters in a single forward pass, the model is deployable on resource-constrained clinical hardware.

As robots increasingly operate in human-populated environments, anticipating human intentions is essential for enabling proactive and socially aware behavior. Automatic anticipation of human-robot interactions is thus emerging as a crucial perception challenge for embodied agents. To this end, we introduce HUI360, the largest dataset for human-robot interaction anticipation in the wild and its set of baselines. The dataset was collected from a mobile robot, in the wild, over multiple days within a 3-month period, and in several environments, capturing natural, spontaneous behaviors from both passersby and users, and encompassing a diverse range of individuals. This variety enables evaluating and improving the generalization capabilities of interaction anticipation models. We designed a pipeline and share code for automatic interaction annotation in arbitrary 360-degree equirectangular videos, along with interfaces for manual refinement. Using this pipeline, we release the HUI360 open set of 1M pre-processed annotations, including detailed 2D poses, facial keypoints, and segmentation masks, obtained using state-of-the-art computer vision methods and manually curated to ensure high-quality tracking and interaction annotation. Additionally, we release the raw panoptic 360-degree images captured from the robot's egocentric viewpoint (on demand, for research purpose only in compliance with GDPR). Finally, we establish benchmark baselines for interaction anticipation, including the first cross-dataset evaluations for this task: to this end, we also release 6M annotations for another existing in-the-wild outdoor dataset collected from a mobile robot (SSUP-HRI). Dataset and code can be found at https://hucebot.github.io/hui360.

A growing class of methods probes a language model by feeding it its own output: self-consistency, iterated refinement, agentic loops. We ask what such a probe measures, in a construction chosen to make the question sharp: a ring of token cells resampled in place by the model's own windowed conditional p_r(x_i | x_{i+-r}). The substrate is Glauber dynamics on token sequences and is not new; what we change is the coupling. Advancing two rings that differ in one token under common random numbers makes undamaged copies diverge by exactly zero, so damage spreading becomes measurable where a maximal coupling gives mixing times instead. The answer is that it measures two different things at once, in readings that look alike. Some quantities are fixed by the construction: the damage light cone is kinematic, and the radius scaling of the token-space Lyapunov exponent lambda_ca(r) is model-invariant across 19 models and two scale ladders spanning 70x. Others genuinely track the model: lambda_ca crosses zero at a reproducible point in training, and the attractor share ranks models consistently however the lattice is built. Left undistinguished, the first kind is readily mistaken for the second -- we did so ourselves for four months, and report a phase transition we measured to three decimal places that belongs to the probe rather than to any language model. We give the test that separates them: hold the construction fixed and vary the model, or hold the model fixed and vary the construction, and see which readings move. We validate the instrument by reproduction first, recovering a Domany-Kinzel damage field bit-exactly against an independent prediction, and we report the estimator failures that this discipline caught -- four retracted verdicts, each on a quantity that looked like a measurement. The methodology ships as a package.

Certified selective predictors attain whatever coverage they attain; operators impose an automation floor: answer at least a $β$-fraction of shifted target traffic with at most an $α$-fraction of answers wrong. Under bounded-ratio covariate shift we prove the Floor Certification Map: once that floor must be certified alongside the selection-conditioned risk $α$, certification acquires a feasibility frontier and a two-resource complexity map, additive up to constants: risk in labeled source, the floor in unlabeled target samples. The rates are local, needing a regular frontier margin, slack below the local-regime threshold, and lattice conditions: pre-registered with a lattice margin for the upper bounds, compatible per-slack for the lower. The displayed split is the operational route; oracle weights also allow a labeled-source floor estimate. Three model-tagged results: a lower bound (Model-B), a matching oracle-weight upper bound (Model-A), and an implementable upper bound (Model-B') valid under a pre-registered exact stratified-shift model with nuisance cost priced explicitly. The match is across these models rather than a single-model minimax theorem, and necessarily so: over the full bounded-ratio class no unknown-weight procedure matches at any sample size (Model-B is inconsistent, witnessed at $α=β=1/2$). The nuisance's necessity is only partially settled. Complexity tracks a localized accepted-region functional, not global effective sample size (ESS), on both sides, though a fixed-ESS separation theorem is left open; both lower-bound axes vanish as $β\to0$, so the floor creates the map. Empirically, the registered bite family diverges with log-log slope $-2.002$ within its pre-registered band; a 1,024-cell audit records 0 violations where the formal certificates fire; and a single-corpus SQuAD-to-NewsQA feasibility audit returns honest refusal.

The Model Context Protocol (MCP) has become the de-facto interface for connecting LLM agents to enterprise tools, and adoption has been explosive: within a year, large organizations went from zero to dozens of internally built MCP servers. That speed created a governance crisis. Each team implemented authentication independently -- some with no auth, some with API keys, some with full OAuth -- producing a fragmented landscape with no consistent way to authorize callers, track who did what, or offboard a departing employee across the fleet. This paper reports an industry deployment that resolves the crisis with a centralized MCP gateway: a single aggregation, governance, and authentication layer that fronts every downstream MCP server. We make four contributions grounded in production experience. First, a two-axis authentication model crossing persona (interactive user vs. automated non-user) with credential type (no-auth, static/dynamic API key, PKCE, client credentials, platform app-context). Second, a gateway authentication layer supporting three enterprise SSO grants and three token-provisioning models: Bring-Your-Own-Token, Generate-Your-Own-Token, and delegated OAuth via RFC 8693 token exchange. Third, three end-to-end identity flows -- User-to-OAuth2, Non-user-to-Service-Account, and User-to-Service-Account -- composing client, gateway, and server. Fourth, the deployment evolution from CDN/WAF/edge perimeter to private MCP tunnels and enterprise-wide connectors. The architecture is in production, fronting dozens of MCP servers across web, desktop, custom-SDK, and low-code clients.

Effective research ideation requires moving beyond a static understanding of prior work to trace how research problems and solutions evolve across the literature. Existing methods either treat papers as unstructured context or model scholarly evolution as isolated citation chains, overlooking interactions among research trajectories. We propose Tree-of-Ideas (ToI), a two-stage framework. EvoTrace reconstructs branching scholarly trajectories from citations, tracking evolving methods, resolved problems, and gaps. EvoAgent then reasons across trajectories to identify convergent problems and complementary solutions, generating grounded research ideas. Across six AI research topics, ToI achieves the highest score among automatic methods (6.27 vs. 5.36 for the strongest baseline on a 10-point scale), with strong Novelty (6.36) and Groundedness (7.00). Also, its score approaches that of human-paper references (6.29), demonstrating the value of cross-path evolutionary reasoning.

Shared memory helps language-model agents reuse information across long workflows, yet relevant evidence may not be admissible for a particular agent or action. Because restrictions propagate through derivations, summaries can conceal private, poisoned, untrusted, or revoked sources, enabling unauthorized reads or unsafe actions. Existing approaches provide semantic retrieval, scoped access, or lineage tracking, but do not clearly separate hard authorization from graded trust or adapt evidence requirements to action risk. We introduce MAP-Graph, a provenance-aware memory layer that represents agents, sources, memories, claims, and actions in a typed execution graph. It traces ancestry, excludes permission-ineligible records, reranks eligible memories by semantic similarity and multiplicative path trust, and applies a risk-sensitive gate before action execution while retaining affected lineage for audit. On a controlled benchmark of 2,700 synthetic tasks per method across three domains, MAP-Graph achieves 94.96% overall task success, 72.70% exact decision accuracy, and 90.22% in the clean setting, where success requires a correct \textsc{Allow} rather than a safe intervention. Ablations isolate the roles of permission filtering, path trust, and action gating, while transfer tests with two additional backbones preserve the exact-decision and access-control advantages. These results support provenance as an operational control signal, rather than only post-hoc audit metadata, within the evaluated setting.

While foundation models have significantly advanced human recognition across diverse modalities, they predominantly rely on static, geometric feature extraction. This approach fundamentally diverges from human perception. Consequently, current models often suffer from "semantic blindness," overfitting to transient noise while failing to leverage invariant soft biometrics, and struggle to capture temporal motion signatures. To bridge this gap, we propose SapiensID 2.0, a human recognition framework enriched with both semantic and temporal awareness. To overcome the lack of soft-biometric annotations, we transfer zero-shot semantic knowledge from Multimodal Large Language Models (MLLMs) into a discriminative embedding space. We resolve the dimensional mismatch between these spaces using Invariant Trait Alignment (ITA) to distill core persistent traits, and Transient Noise Disentanglement (TND) to decouple artifacts like clothing. Furthermore, we design a Kinematic Semantic Attention Head (K-SAH) that extends spatial attention across temporal windows. By tracking semantic patches over time, K-SAH captures rich kinematic signatures without requiring large-scale video datasets. Extensive experiments demonstrate that SapiensID 2.0 achieves state-of-the-art performance across image- and video-based person re-identification and gait recognition, while maintaining robust face recognition capabilities.

Complex software systems develop over timescales that exceed the lifespan of any individual coding agent. Most agentic software systems preserve continuity through persistent sessions, memories, managers or shared context. We introduce EvoX Genesis (hereafter, Genesis), which instead makes the software project persistent while allowing local agents to remain finite-lived. Genesis represents software as a persistent recursive world: each local world is situated by an accepted version and a repository path, finite-lived agents propose local changes, recursive delegation moves work across paths, and only accepted consequences advance the persistent version history. We evaluate this organization across formation, continuation and redevelopment. Starting from a repository with no compiler implementation, Genesis used DeepSeek V4 Flash to build a Rust-based C compiler with about 250k tracked lines; the run lasted over 120 hours, archived over 1,000 agent episodes and incurred only US$44 in model-token charges. The compiler passed the complete c-testsuite and most LLVM and Csmith tests. In a separate compiler world generated with GLM 5.2, development continued after repeated agent replacement while retaining full test performance. Genesis also reimplemented 13 MESA modules with over 100k Fortran lines as a Rust workspace with nearly 90k Rust lines; across six numerical workloads, it achieved median speedups of 1.55--6.87x. These results show that long-horizon software development can be organized around a persistent project rather than a persistent agent.

Reasoning shortcuts are solutions of a neurosymbolic system's rules that produce correct predictions through unintended concepts. A recent framework of Takemura, Inoue, and Nishino analyzes them through an automorphism group of value relabelings and asks, as its central open question, when rules pin concepts down. We first show that the framework's key definition, one shared permutation applied at every position, does not apply as stated to any of the four heterogeneous benchmarks it was evaluated on, and that the most direct embedding, padding domains to a common size, produces confident false pathology: 90.91% of solution pairs reported unexplained on CLE4EVR, where every well-defined member of the hierarchy we introduce reports 0%, and the padded verdict's content rotates with configuration-file ordering. Re-measuring eleven rule families under fifteen pre-specified predictions (thirteen confirmed), unexplained-pair rates span 0% to 99.9999% and track provable structure: six theorems give sufficient conditions for transitivity and its failure, including a Free Slot Lemma certifying Kandinsky's pathology from syntax alone. For circuit-given rules, deciding symmetry-inertness of a coordinate is coNP-complete; nontrivial-automorphism existence is coNP-hard under randomized reductions, lies in $Σ_2^p$, is not $Σ_2^p$-complete unless PH collapses, and on monotone circuits is coNP-complete outright. In the Boolean case transitivity is classified exactly: automorphisms explain everything iff the solution set is an affine coset. Weakly supervised models place all 94 observed shortcuts at the one level the componentwise theory flags and none at the 48 it certifies transitive; twelve typed-ambiguous levels produce none, separating what symmetry permits from what optimization selects, and a dual-head control replicates the geography. All numbers trace to released artifacts.

Direct teleoperation of vehicles faces critical technical bottlenecks: communication latency and the operator's inability to physically perceive unmodeled environmental disturbances (e.g., aerodynamic drag, bank angles) coupled with highly nonlinear tire-road dynamics. To address these challenges, we propose a tailored unilateral teleoperation framework. The system integrates the Wave Variable (WV) approach to passively guarantee stability under stochastic delays, and an adaptive Radial Basis Function Network (RBFN) to actively compensate for vehicle-specific uncertainties. Unlike existing WV-neural network architectures designed for bilateral robotic arms, our framework features decoupled adaptive laws specifically designed for vehicle longitudinal and lateral dynamics. Furthermore, compared to model-heavy predictive controllers, the model-free RBFN offers rapid online adaptation without heavy computational overhead. Building upon our preliminary theoretical formulation, this brief paper presents comprehensive comparative analyses and real-world hardware validations. Simulation benchmarks against PID, LQR, MPC, and NMPC demonstrate that the RBFN achieves superior robustness against unmodeled disturbances while requiring orders of magnitude less execution time than MPC and NMPC, making it ideal for resource-constrained vehicle edge computing. Finally, hardware-in-the-loop experiments using a 1/10th scale vehicle over a 4G network validate the system's practical feasibility, safety, and robust trajectory tracking under physical road uncertainties.

Competitive artificial-life systems can rank trained controllers differently under training and ecological evaluation. We present Neuroevolution Arena, a GPU-accelerated spatial ecology of independently parameterized neural-network cells, and an audit-tracked nested evaluation protocol. Three implementation-specific update-and-inheritance regimes (EvoEvo, EvoRL, and RLRL) are crossed with two neural architectures for 50,000 generations in three independent training runs per condition. One saved elite-controller artifact from each of the 18 runs enters an aligned-run frozen-evaluation design comprising 198 computational jobs. Pairwise effects average three seed-defined ecological contexts (two cooperation-permitting and one attack-permitting) within each aligned training-run block; the independent level remains n = 3 runs per condition. RL-enabled regimes attain higher recorded training fitness than EvoEvo, whereas pairwise outcomes show architecture-conditioned majority patterns and substantial artifact dependence. Six-way winners vary across artifacts and contexts, and the prespecified survival endpoint has a complete floor. We contribute a nested protocol that separates training-run artifacts from evaluation contexts and exposes, rather than conceals, their different sources of variation.

Detecting a traffic anomaly does not establish whether a video-language model can explain what happened, localize it in time, or identify its causes. We introduce TAR (Traffic Anomaly Reasoning) and TAR-Bench, paired resources for training and evaluating these complementary capabilities across 10 tasks spanning question answering, temporal reasoning, and scene understanding. TAR contains 44,040 automatically generated annotations with chain-of-thought traces for 3,670 CCTV videos from eight public datasets. TAR-Bench provides 960 human-curated annotations for 80 held-out clips from 17 public YouTube videos. Evaluation of eleven vision-language models reveals a gap between question-answering performance and temporal or scene reasoning. Progressively adding task groups during supervised fine-tuning improves aggregate performance on both Cosmos-Reason2-8B and Qwen3-VL-8B-Instruct. Training on all 10 tasks raises their mean benchmark scores from 34.3 to 55.7 and from 30.9 to 53.9, respectively. These results support joint training across complementary tasks as a promising approach to traffic anomaly understanding, while highlighting persistent limitations in temporal precision and causal attribution. TAR and TAR-Bench serve as the official training and in-domain evaluation resources for AI City Challenge 2026 Track 3.The dataset is available at https://huggingface.co/datasets/nvidia/PhysicalAI-Traffic-Anomaly-Reasoning

Accurate vessel trajectory prediction is critical for maritime safety and anomaly detection, yet existing models often struggle with geographic bias and navigational realism. We propose the Continuous Regression Hybrid Transformer (CRHT), a deep learning framework designed to forecast vessel motion using Automatic Identification System (AIS) data. To mitigate spatial data imbalance, we introduce an online K-means cluster sampling strategy that ensures diverse exposure to rare maneuvers during training. Our hybrid architecture integrates 1D convolutional layers for local kinematic feature extraction with a multi-head attention mechanism for global temporal context. CRHT demonstrates superior performance in short-term forecasting, achieving the lowest errors at the 1-hour horizon. The results demonstrate that while discrete models provide high navigational stability over long horizons, CRHT offers an optimal balance of precision and maneuver tracking for real-time maritime surveillance.

Humanoid locomotion in highly confined environments requires navigating dense environmental obstacles and complex self-collision bounds while maintaining multi-contact dynamic feasibility. Traditional trajectory optimizers frequently struggle in these restricted spaces, as navigating the large collision space with splines on particle abstractions is insufficient and leads to poor local minima. To address this, we propose a three-stage whole-body planning framework that formulates kinematic path planning directly over kinematically reachable rigid-body volumes. By integrating differentiable collision avoidance into a reachability-constrained formulation, our framework synthesizes volume-informed guides that reliably guide a full-order trajectory optimizer over long horizons. We show that these optimized plans serve as high-quality references to train a residual reinforcement learning policy for robust online execution. We validate our approach on the Unitree G1 humanoid across three benchmark testbeds exceeding NIST emergency response standards, achieving restricted confinement ratios ($C_r < 1.5$). Our framework generates feasible trajectories across 12-to-18-second tasks with complex foot and hand contacts where standard baselines fail, while the learned policy successfully tracks these plans under extensive domain randomization in physics simulation.

Safe reinforcement learning maximizes reward subject to safety constraints. For Constrained Markov Decision Processes, the linear-programming view over occupancy measures implies that whenever the constraint is active at optimality, the optimal policy lies exactly on the constraint boundary, yet standard gradient-based methods do not exploit this structure and often settle in the feasible interior. We introduce Boundary-Seeking Policy Gradient (BSPG), a first-order method whose update combines a tangential component that improves reward while preserving cost to first order with a signed, residual-driven normal component that regulates the policy toward the active boundary from either side; the combined direction admits an algebraic Lagrangian form with an induced coefficient and no learned dual variable. Under exact gradients and stated regularity conditions, the constraint residual converges to zero from either side with a finite-horizon $O(1/\sqrt{T})$ bound, the tangential component is a reward-ascent direction on the boundary, and any convergent parameter sequence is stationary on the active constraint set, satisfying the KKT conditions when the limit is also a local maximizer over the feasible set. This complements existing analyses, which certify feasibility but do not characterize the constraint value at convergence. On a standard Safety-Gymnasium navigation task, BSPG attains higher reward while tracking the boundary more tightly than the compared baselines.

High-performance Tensor Core kernels rely on a low-level PTX pipeline built from asynchronous data movement with cp.async, warp-level matrix loads with ldmatrix, and matrix multiply-accumulate operations with mma.sync. However, most application code accesses Tensor Cores indirectly through the WMMA C++ API. This paper asks a focused, practical question: when does replacing WMMA with hand-written PTX actually pay off? To answer this question, we conduct a controlled, single-GPU study on an NVIDIA L4 GPU (Ada, SM89), comparing double-buffered WMMA baselines with a family of hand-written PTX GEMM kernels across FP16, INT8, and INT4 arithmetic and square problem sizes from $N=512$ to $N=8192$. Every kernel is profiled with Nsight Compute across the full metric set, and PTX speedups are reported relative to the corresponding same-precision WMMA baseline. Hand-written PTX provides no end-to-end speedup for FP16, because its instruction-level gains are offset by operand-packing overhead. In contrast, the PTX kernels achieve consistent speedups of 1.4x-1.8x for INT8, driven primarily by lower instruction counts and better global-memory coalescing, and 2.9x-4.3x for INT4, where native mma.sync.m16n8k64.s4 execution avoids the software-emulated sequence used by the WMMA path. Relative to the FP16 WMMA baseline, the best quantized kernels reach 34.4x (INT8) and 98.7x (INT4) at $N=8192$. Across these experiments, occupancy is a poor predictor of throughput. For large matrices, performance instead tracks memory-system behavior -- particularly global-load coalescing and DRAM-active cycles -- more closely than Tensor Core utilization. These results identify the precisions and operating regimes in which the additional complexity of hand-written PTX is justified.

Bioacoustic recordings are often degraded by ambient noise, which complicates the analysis of weak or noise-overlapped vocalizations. Convolutional neural networks, particularly U-Net architectures, have shown a strong denoising performance in speech and music processing. However, their direct application to bioacoustic signals is limited by the scarcity of clean training data. To address this issue, we propose a training set synthesis approach and develop a supervised denoising model that predicts a complex ratio mask in the time-frequency domain. The model leverages ridges, or frequency contours, that represent the fundamental frequency together with one or more harmonic partial components of vocalizations. These ridges are used both for the synthesis of training sets and to design a loss function that assigns higher weights to the ridge regions (ridge-guided loss function). This weighting step helps the network better preserve vocalization details during denoising. As a case study, we evaluate our approach using ultrasonic vocalizations (USVs) recordings of house mice, which are widely studied in behavioral biology and neuroscience. In actual field recordings, the proposed method enhances fundamental and harmonic partial ridge tracking compared to our previous signal-processing approach. In addition, a classifier trained on denoised data improves USV classification on out-of-sample, noisy recordings from wild and domesticated mice compared to classifiers trained on noisy recordings. Our proposed method also substantially improves the scale-invariant signal-to-distortion ratio on synthetic testing data across a wide range of input signal-to-noise ratios. Although we focus on USVs, the proposed approach should be broadly applicable to other bioacoustic signals with trackable ridges, and thus enables ridgebased training set synthesis and denoising.

Muon's polar update is exact for an unweighted spectral geometry. We introduce GO-MUON, which uses a matched data-dependent geometry and reuses it across several optimization steps. Conditioned on any positive-definite left and right maps, its raw update exactly solves the corresponding weighted spectral oracle; this statement is independent of how the maps are estimated or how recently they were refreshed. For softmax cross-entropy, we quantify when the observed-label backward factor approaches the model Fisher and generalized Gauss--Newton factor. We also show that four-step refresh nearly preserves the tracking delay of slowly changing geometry while increasing stationary factor noise, making lazy geometry a compute--statistics tradeoff rather than a denoising mechanism.

Graph Neural Networks (GNNs) suffer from two fundamental limitations: over-smoothing, where node representations become indistinguishable with depth, and over-squashing, where long-range information is compressed through limited message-passing channels. Existing metrics such as Dirichlet energy provide global characterizations of over-smoothing but lack the resolution to analyze node-level behavior and guide architectural improvements. In this paper, we propose LEED (Local Embedding Evolution Distance), a novel local metric that quantifies over-smoothing by tracking the evolution of individual node embeddings across layers. By operating at the node level, LEED enables fine-grained analysis of representation dynamics during training, revealing heterogeneous over-smoothing patterns that are invisible to global energy-based measures. This locality induces informative node importance scores, interpreted as embedding-driven centrality measures. We leverage LEED to design a more efficient strategy for virtual node selection. Unlike existing approaches that depend on multiple heuristic centrality measures, our method uses LEED as a unique criterion to guide the construction of Local Virtual Nodes to mitigate over-squashing. Experiments show that LEED provides more informative diagnostics than Dirichlet energy while preserving global evaluation, and enables more effective virtual node integration, improving GNN performance across datasets.

Connected Autonomous Vehicles (CAVs) increasingly exploit Vehicle-to-Everything (V2X) communication to exchange multi-source sensor information, enabling advanced Collaborative Perception (CP) capabilities. Extending beyond these capabilities, this work focuses on Collaborative Joint Perception and Prediction (Co-P&P), a paradigm that unifies CP with motion prediction to mitigate two persistent challenges: the accumulation of perception errors and visual occlusions. We present a conceptual framework for Collaborative Joint Perception and Prediction (Co-P&P) that improves motion prediction of surrounding road users, thereby enhancing situational awareness in complex and dynamic traffic environments. Building upon our preliminary study, this extended version compares the performance of different fusion strategies and establishes baseline performance for a modular design of perception and prediction. Experimental results show that prediction-level fusion leads to a decline in overall system performance compared to detection-level or tracking-level fusion. We further implement a minimal end-to-end Co-P&P prototype that couples collaborative point-cloud sharing via the RENO neural codec with joint detection-forecasting via FutureDet, showing that collaboration improves forecasting accuracy while neural compression preserves this benefit at roughly 34x lower communication bandwidth.

Long-horizon future-frame prediction is important for autonomous driving, traffic surveillance, and intelligent transportation systems, yet remains challenging due to temporal ghosting, geometry drift, and inconsistent object motion. Recent latent video diffusion models have achieved impressive visual quality, but directly applying them to structured traffic scenes often leads to unstable geometry and degraded temporal coherence over extended horizons. We present a training-free inference framework that stabilizes reliable static structure in pretrained video predictions through multi-frame temporal context and view-conditioned routing. For front-camera videos, our method refines generated futures with a multi-frame depth-layered renderer that projects static geometry from observed history frames while preserving dynamic regions from the generative base model. For heterogeneous traffic views, a frozen vision-language model infers a coarse camera group from the observed clip and selects a specialized motion-based predictor. The framework requires neither retraining nor fine-tuning of the underlying video model and can be applied directly to pretrained generators. We validate the proposed framework on the AI City Challenge Track 5 benchmark, where our final system achieves competitive performance among the top-ranked teams. These results demonstrate that geometry-aware inference-time refinement and view-conditioned hybrid inference can improve static-geometry stability and low-level structural fidelity without changing the original model architecture.

Simulation is crucial for developing and testing autonomous driving systems. In particular, the development of localization and control algorithms relies on an accurate vehicle dynamics simulation. However, most vehicle dynamics models are two-dimensional while real-world roads are three-dimensional. For example, effects from the three-dimensional road geometry on the Las Vegas Motor Speedway can increase the normal forces on the tires by more than 66% compared to the nominal load at standstill. As a result, even highly detailed planar vehicle dynamics models struggle to accurately reproduce the real vehicle's behavior. While solutions for three-dimensional vehicle dynamics exist, they are rarely adopted, computationally expensive, and complex. To address this issue, we present a novel method to couple planar vehicle dynamics models with real-world three-dimensional road geometry. We transform the planar vehicle state from the vehicle model's two-dimensional plane to its corresponding representation in three-dimensional space. Additionally, we calculate road-geometry-induced forces and moments and apply them to the planar vehicle model. We validate our approach using high-speed data recorded with a full-scale race car on the banked Las Vegas Motor Speedway. Furthermore, on synthetic tracks, we show that our method yields accurate results even in edge cases. Together, our results demonstrate that the gap between planar simulation and real-world three-dimensional roads can be closed without abandoning simpler planar models. To simplify adoption of our method, we provide the implementation as open-source software on github.com/TUMFTM/3d-road-geometry-coupling.

We identify and quantify temporal misgrounding: the systematic retrieval and citation of the currently in-force version of a legal article when the applicable version is an earlier or future one. Standard legal RAG treats the corpus as static; we argue legal question answering is a temporally-indexed retrieval problem. We introduce FiscalQA Pro, pairing a versioned corpus of 32,436 article-versions of the French tax code (93 years, 1938-2031) with an all-model-hard temporal-reasoning track: 209 scored, expert-reviewed questions across 33 CGI articles (221 released; twelve flagged out of the answerable scope). At selection time, no evaluated model recovered its date-applicable answer closed-book in any of four sampling draws, and the currently in-force text lacks the gold value for all but one of the scored questions. Answers are scored deterministically via atomic ground-truth "nuggets" (regex and numeric-with-tolerance), never LLM-as-judge: an LLM judge would inherit the temporal bias it is meant to score. Across eleven models (five frontier closed-API systems plus Gemini 2.5 Pro as a substitute entry, and five open-weight), parametric knowledge yields 3.0% mean strict accuracy and RAG over a static current-version corpus 2.7%. Static RAG retrieves the date-applicable version 0% of the time, confidently citing a real but inapplicable version. Our end-to-end retriever over a multi-version index, with no oracle, reaches 98.3% mean strict; an oracle-article ablation reaches 99.1%, locating the residual gap in first-stage recall, not version selection. We additionally release a version-aware jurisprudence dataset of 69,208 citation links, together with the corpus, benchmark, model responses, and pipeline code.

A CAD model is not engineering-grade merely because it looks correct. It must satisfy design requirements, respond predictably to parameter changes, support controlled edits, match a reference structural response under a declared analysis, and connect to other parts through valid joints. We present CADEngBench, a two-track benchmark for these capabilities. CADEngBench-P evaluates 300 parametric parts, each used for one zero-to-CAD task and one functional-editing task (600 tasks in total), through boundary-representation (B-Rep) validity, engineering and DFM checks, parameter-family perturbations, functional editing, and matched linear-static FEA in CalculiX. CADEngBench-A evaluates 150 body pairs through ranked joint retrieval, exact face-and-edge grounding, joint-frame prediction, and kinematic verification. Across eight multimodal, code-capable models, editing supplied CAD is substantially easier than generating it, while complex edits and matched FEA remain difficult. Assembly predictions often locate the relevant region but fail to recover the recorded joint or mating entities. These results show that CAD evaluation must test engineering behavior rather than appearance alone.

Outcome verifiers score completed reasoning traces but do not assign credit to intermediate tokens. Privileged self-distillation attempts to fill this gap by rescoring a model's own rollout with training-only information. A token likelihood change, however, is not automatically outcome credit. We separate three questions: whether the score tracks better actions, whether feedback construction changes what is compared, and what behavior the training loss reinforces. We establish these distinctions formally. When a rollout is scored using hindsight feedback written about that same rollout, its content determines both the tokens and the scoring context, creating direct self-dependence. Using feedback from another rollout of the same problem removes this dependence but does not guarantee a useful score. In matched experiments with a 20B model on AIME 2025, the implemented additive score is near chance (AUC=0.505) and slightly favors incorrect traces after length adjustment. In the paired comparison, the outcome-only control records 64.2%, versus 24.2%--33.9% for five token-score variants. The results motivate validating score meaning, feedback construction, and training behavior separately before calling a likelihood signal credit.

LLM analytics agents are evaluated on SQL syntax accuracy, but production failures look different: questions with two valid business definitions, questions the warehouse cannot answer, deprecated columns after a schema change, and queries that execute successfully while returning the wrong business number. No execution-match metric can score them. This paper introduces WarehouseReliabilityBench, 400 frozen tasks over two synthetic warehouses in which roughly half the correct responses are a clarification, an abstention or a refusal, with pinned denominators and a pre-registered paired bootstrap fixing each claim verb before the numbers existed. QueryProof, a 7B agent, uses rules derived from a semantic layer and physical catalog to determine its behaviour, and gates every answer on deterministic post-execution checks. On an 80-task synthetic test split evaluated once, QueryProof outperforms a direct-prompted 32B baseline by +0.237 [+0.112, +0.375] Business Truth Rate at 71.0% lower cost per correct answer; against a cost-matched few-shot baseline the accuracy gain holds but the cost difference does not resolve. This compares systems rather than model sizes: the 32B baseline receives none of the scaffolding. False success falls from 0.754 to 0.351 of returned answers, and no wrong number was returned on an answerable task (0 of 24), though 13 answers went to questions requiring clarification or abstention. Removing the routing layer changes little (0.562 against 0.537), so the result does not depend on escalation. Routing tuned on validation over-abstains on test, and the fitted confidence model loses to the heuristic it replaced. Resampling template families rather than tasks widens both accuracy intervals to include zero, so the effect's direction is better supported than its magnitude. The gain tracks the deterministic layer, though no component ablation was run.