Science & Mathematics

EEG foundation models are advancing at an unprecedented pace, yet their evaluation is driven by downstream accuracy, which reveals little about whether their predictions rely on physiologically interpretable signal structure. We introduce a set of perturbation probes in two grounding axes: spatial grounding in channel identity and topographic locality, and spectral grounding in oscillatory content and waveform morphology. We apply these probes to five published EEG foundation models across motor-imagery and seizure recordings. Our experiments reveal differences in both representational sensitivity and downstream reliance. Electrode identity may be encoded without influencing downstream classification: permuting LaBraM's positional codes heavily displaced its embeddings yet reduced motor-imagery accuracy by only 1.2 points, while spectral details may be discriminable without producing specific responses to simulated clinical markers, as only REVE exceeded a spectral baseline on this task. These dissociations are hidden in common benchmarks, demonstrating how grounding probes can complement performance-based evaluation by revealing what EEG foundation models encode and what their predictions depend on.

Key-value (KV) cache capacity and memory traffic are primary bottlenecks in long-context autoregressive inference. We evaluate five cache paths on an Apple Silicon M5 Max using MLX: a native bf16 baseline, block-scaled E2M1 FP4, a Python reference implementing radial quantization with a QJL-style residual, sequential FP4-to-Polar composition, and a fused Metal Polar kernel. On Apple Silicon M5 Max, fused Metal PolarQuant achieves 46.8 tok/s at 3,456 B/token - 3.64x compression versus the bf16 baseline and 1.84x faster than the Python Polar reference (25.4 tok/s), while slightly outperforming block-FP4 (41.5 tok/s at 3.62x compression). Sequential composition of FP4 and Polar (Double Optimization) yields no additive benefit: 4,241 B/token at 15.1 tok/s, worse than FP4 alone on both axes. The fused path establishes that radial coding can be competitive when packing, rotation, and reconstruction are placed in one device kernel, but this result does not extend to longer contexts. At 32K and above, all compressed variants collapse to 1.5-14 tok/s because the current attention path repeatedly reconstructs the full cache, retaining cumulative O(N^2) decode work. At 64K, the Metal variant runs out of memory when a 33.5 GB allocation exceeds MLX's 22.6 GB Metal buffer limit. Fused scaled dot-product attention operating directly on compressed KV, together with chunked pre-allocation, is required for long-context viability. These results separate representation-level compression from kernel and allocator effects and identify the conditions under which low-bit KV caching helps on unified memory hardware.

Robots deployed in human-centered environments need maps that support both geometric navigation and open-vocabulary semantic queries beyond a fixed label set. A key challenge is enriching such maps with semantic information without relying on manually annotated segmentation masks or closed-set semantic categories. Existing geometry-driven active mapping maximizes coverage but treats semantics as a passive by-product of the collected trajectory, whereas active semantic mapping typically depends on supervised semantic segmentation. We present AOV-GS, an active 3D Gaussian mapping system that guides exploration using frozen foundation models. During coarse exploration, AOV-GS re-ranks geometrically promising candidate views according to CLIP mask-embedding novelty computed from SAM instance segmentation, biasing exploration toward semantically informative regions. Frozen SAM and CLIP provide instance masks and language-aligned embeddings throughout mapping, while a lightweight scene-specific language field is fitted offline directly to the collected CLIP embeddings rather than manually annotated masks, enabling open-vocabulary text-query segmentation on novel views. In a controlled comparison using the same SAM-CLIP perception stack, AOV-GS improves average language-field mIoU over geometry-only active planning by 1.3 percentage points on Replica (34.6% on eight scenes) and by 1.8 points on Matterport3D (23.1% on five scenes). With CorrCLIP, AOV-GS reaches 35.6% on Replica, 2.0 points above the matched Geometry-based + CorrCLIP baseline. Novel-view PSNR remains comparable to the geometry-only baseline on both datasets while depth L1 improves on Replica. Code is available at https://github.com/OptimusFaber/AOV-GS.

Social-media behavior can be observed through both interaction structure and multimodal content, yet each view is often incomplete. We study two complementary tasks on separate datasets: predicting retweet links from an extremely sparse graph without node attributes, and generating user intentions from text-image posts with limited annotations. For link prediction, we use a leakage-controlled transductive protocol to compare GCN and GraphSAGE, classical topology heuristics, and three node-input designs. Across repeated runs, a GCN with compact degree and strength features achieves an average precision of 0.9243 and an MCC of 0.7051. Under joint edge-label corruption, topology-based GCNs retain strong ranking performance, whereas fixed-threshold MCC degrades substantially; adding fixed random node identifiers does not improve overall robustness. For intention recognition, we retain the MIKO framework and evaluate LoRA adaptation of three language models together with distribution-aware sampling based on Pearson correlation, cosine similarity, and KL divergence. Fine-tuning consistently improves lexical-overlap metrics, while distribution-aware sampling produces more uniform performance across intention types. Together, the two studies characterize who interacts and why users post when social-media observations are sparse, noisy, or limited.

Learning-based path planners based on diffusion and flow matching can generate diverse trajectories from demonstrations but typically lack mechanisms for constraint satisfaction and collision avoidance during deployment. This work proposes a framework that combines flow-matching-based path planning with model predictive path-following control to integrate data-driven path generation with real-time, constraint-aware execution. The flow-matching model captures multimodal path distributions, while the predictive controller adapts motion online to satisfy state and input constraints and avoid obstacles using updated local sensor information. An event-triggered re-planning scheme incorporates predictive control solutions into subsequent path generation, promoting collision-free navigation in environments with previously unseen obstacles. Simulation results for mobile robot navigation demonstrate robust performance across cluttered environments. The framework provides a systematic approach to combining flow matching with predictive control for adaptive, data-driven navigation.

Quantization-aware training (QAT) aims to recover reasoning capabilities in low-bit large language models (LLMs). Existing recipes typically supervise the quantized student with the original full-precision model on fixed response sequences. However, the resulting low-bit quantized models still struggle on long-horizon reasoning tasks and often fail to terminate within a token budget that suffices for the full-precision model. We identify two challenges in closing this gap. First, fixed-response supervision leaves a mismatch between training prefixes and those encountered during student generation, and this trajectory divergence grows as responses lengthen. Second, on its own prefixes, the student underpredicts some next-token candidates and overpredicts others, and this distributional misalignment becomes more frequent at high-entropy prefixes. We present OnQ, a low-bit recovery pipeline that combines on-policy distillation and candidate-wise Dynamic KL. Starting from block-wise calibration, the quantized student generates responses and receives full-precision teacher supervision on its own prefixes. Dynamic KL weights forward and reverse KL by each candidate's teacher-student probability ratio to balance correction of under- and overprediction. Across Qwen3-1.7B, 4B, and 8B at 2- and 3-bit precision, OnQ achieves the highest nine-benchmark average, with the largest gains on long-horizon reasoning benchmarks. At 2 bits, it outperforms every baseline on every benchmark at every model size, improving the Qwen3-8B average by 10.26 points over the strongest baseline. It also produces shorter responses with fewer truncations and repetitive errors.

LLM-based vulnerability discovery harnesses can miss real, reachable vulnerabilities because of how the harness around the model is built. Steps that select code, filter it, and store and merge findings can drop a vulnerability before or after the model runs, independently of the model's capability, and the report usually gives no sign that this happened. These drops come from engineering choices, and errors. They leave vulnerabilities undetected, and an attacker who knows the harness's rules can also use them to hide a vulnerability on purpose. Understanding these drops shows how to build harnesses that fix or disclose them and are harder to evade. Starting from the two harnesses we maintain, Knostic's OpenAnt and raptor, and then examining 10 additional open-source harnesses, we mapped 13 points where such drops occur and organized them into a taxonomy. Using each harness's own code, we show a single file that six harnesses dropped before their model saw its code, five of them without notice. We propose engineering failure patterns mapped to these drops, and relate them to Ptacek and Newsham's work on intrusion detection evasion. The paper also covers content-steered verdicts and prompt injection, and gives design principles and tests for harnesses that report accurately what they analyzed. The reliability findings likely apply to other LLM harnesses, beyond vulnerability research and application security, and the evasion findings apply wherever an attacker can write or influence the code being checked. We also argue that harnesses built to be descriptive can, in effect, be prescriptive. Rules written to describe typical code and typical model replies end up deciding what is analyzed, and anything that does not conform is treated as absent or as "no vulnerability."

Practical image generation requires both human preference alignment and few-step efficiency, yet combining reinforcement learning (RL) with distillation is often multi-stage and may erode RL gains We observe that few-step RL progressively improves structural coherence, even with specialized rewards, but leaves low-level artifacts that these rewards may fail to penalize. To address this gap, we propose AdvFidelity, which integrates artifact-sensitive fidelity supervision into RL through an online adversarial reward. This enables joint reward alignment and high-fidelity few-step generation within a single RL stage, without distribution matching distillation (DMD) or other explicit distillation objectives. We train a discriminator to rank high-quality references (positives) above content-matched reconstructions (negatives) produced by the current generator. Reconstructing moderately noised references largely preserves style and scene layout while exposing low-level degradations. This encourages the discriminator to assess artifacts rather than content differences. With negatives reconstructed online, alternating updates adapt fidelity supervision to the generator's evolving failure modes. References can come from existing target-domain data or full-step model samples. Built on LeapAlign, AdvFidelity backpropagates gradients from both rewards through the few-step sampling trajectories used at inference. Across two model backbones and three task settings, AdvFidelity matches full-step LeapAlign in alignment, task performance, and overall generation quality and remains highly competitive with other evaluated full-step RL methods, using only 4 or 8 steps without CFG.

We propose a novel stochastic cascading bandit setup where probing an ordered shortlist of $K$ out of $L$ arms incurs position-dependent costs, and the learner observes only noisy feedback regarding each arm's true output. We identify a sub-class of noise models for which the optimal shortlist can be found by sorting arms according to a per-arm score: such a score exists for both selection and ordering when the channel has no false positives, and for ordering alone (but not selection) when measurements are costless; however in general neither selection nor ordering is separable once noise and cost are both present. For the case where both selection and ordering scores exist, we provide a UCB-based algorithm with $O((L-K)\log T)$ instance-dependent regret over a time horizon $T$, which matches an algorithm-agnostic lower bound (up to constant factors). For the more challenging scenario where scores don't exist, we develop a new algorithmic template, DP-UCB, that embeds optimism directly inside an exact dynamic program, yielding $O\big(LK\log T\big)$ regret. Finally, we validate our theoretical results through extensive numerical evaluations and simulations over both synthetic and real-world datasets.

Radio maps characterize the spatial distribution of wireless conditions and are fundamental to network planning and optimization. However, existing learning-based methods reconstruct isolated radio-map snapshots and do not model how network actions drive the evolution of the radio environment. In this paper, we present RadioDreamer, a two-stage world model that unifies action-conditioned radio-map forecasting and imagination-based network optimization within a latent space. To learn causal environmental transitions from sequential interactions, we introduce a structured spatio-temporal transformer backbone coupled with dynamic-region supervision, enabling accurate modeling of action-sensitive areas. Second, rather than relying on conventional physical trial-and-error, RadioDreamer establishes a general planning paradigm that generates counterfactual latent trajectories to evaluate and optimize long-horizon network parameter configurations. Experiments show that RadioDreamer improves both multi-step rollout forecasting and closed-loop network optimization. It reduces full-map forecasting error by 25.2% on average and achieves approximately 80% accuracy in predicting the evolution direction of radio signals. More importantly, for long-horizon network configuration, imagination-based planning surpasses model-free policies optimized directly on radio-map observations by 8.8% in accumulated return.

Language models relying on verbalized Chain-of-Thought (CoT) are constrained to token-by-token generation in the language space, where many tokens serve textual coherence rather than core reasoning. This entangles genuine differences in logical trajectories with superficial variations in verbalization. Moving reasoning into a compact, continuous latent space offers a promising alternative for efficiently representing and exploring diverse reasoning trajectories. However, current latent models remain largely deterministic, capturing only a single reasoning trajectory rather than modeling a distribution over multiple valid alternatives. While stochastic approaches parameterize distributions over latent reasoning steps, they frequently suffer from unstable optimization and limited empirical performance. Furthermore, their reliance on autoregressive latent prediction during training precludes effective parallelization, leading to prohibitively long training times. To overcome these bottlenecks, we propose Probabilistic Latent Reasoning Models (PLRM), a parallelizable variational framework for probabilistic latent reasoning. PLRM adopts a JEPA-style objective in which the posterior infers latent distributions in parallel from observed reasoning trajectories, while the prior learns to predict the subsequent latent distribution via teacher forcing. To model heterogeneous reasoning processes, we introduce a discrete strategy variable together with a mixture-based prior, allowing the model to preserve the Gaussian-mixture structure induced by multiple distinct reasoning trajectories rather than collapsing them into a single Gaussian. Across diverse benchmarks, PLRM achieves performance competitive with deterministic latent reasoning models on single-trajectory datasets while enabling efficient training and consistently outperforming competing stochastic baselines on multi-trajectory datasets.

The atomic ordering in Fe-rich Fe-Si alloys strongly influences their magnetic properties and mechanical behavior. While general features of long-range atomic ordering are known from the experimentally established phase diagram in this composition range, detailed atomic configurations and driving forces behind them remain speculative. Here, we address this gap by using effective cluster interactions (ECI) obtained from density functional theory (DFT) calculations and subsequent statistical thermodynamics modeling to unravel the interplay between magnetism, composition, and temperature in body-centered cubic (bcc) Fe-Si alloys up to 20 at% Si. Our analysis reveals that the ordering behavior is governed by a strong coupling between magnetic order, lattice expansion, and Si concentration, with specific ECI shells switching sign or magnitude as the magnetic state changes, which is reflected in the phase equilibria and ordering behavior of Fe-Si alloys. In particular, we demonstrate that D0$_3$ ordering is, in fact, primary in the totally ordered ferromagnetic state, while the B2 type of ordering is promoted in the paramagnetic state at elevated temperatures. The temperature-dependent evolution of these ordering patterns, obtained from Monte Carlo simulations based on the magnetically resolved ECIs, reproduces the experimentally observed phase equilibria and short-range order characteristics. By explicitly linking the magnetic state with the effective interactions and macroscopic phase behavior, this work establishes a predictive framework for an ab initio-based description of ordering in Fe-rich Fe-Si alloys and highlights the decisive role of magnetism in ordering phenomena in this system.

Public descriptions of the Starlink downlink specify synchronization sequences, edge pilots, and predictable symbol templates, but leave the coded control fields and explicit payload-code representations largely unresolved. We examine one continuous complex-baseband capture spanning approximately 24 ms and recover control and coding structure beyond those predictable elements. Reference-phase removal exposes terminated rate-1/3 convolutional control blocks with constraint length seven and a group-level cyclic redundancy check. A conditional descriptor layout predicts sequential allocations through per-codeword symbol rounding, agreeing exactly with independently estimated endpoints in 16 of 18 frame records and within two symbols in the remaining records. Quasi-cyclic closure, binary elimination, and sparse dual checks yield explicit systematic code representations at rates 1/2 and 5/8. Frozen encoders predict 78,848, 9,984, and 811,008 reserved parity bits for three formats without disagreement. A rank-deficient long-code fit is completed within a declared finite structural family, and shared encoders are tested across modulation formats. The recovered representations explain observed coded symbols; they do not authenticate transmitter-standard identities or plaintext semantics. Two high-order formats have only one target field each, while codes 28 and 37 admit verified shortened cores with unexplained extra bits and code 4 remains unresolved.

3D semantic occupancy prediction estimates voxel-wise occupancy and semantic labels throughout a scene. Our experiments reveal that conventional sparse transposed-convolution upsampling yields poorly differentiated voxel responses and that successive decoder layers discard previously correct semantic predictions. To address these issues, we present Feature-Diverse Upsampling and Memory-Aware Decoding for 3D Occupancy Prediction (FM-Occ), comprising two components. First, 3D Sparse Subvoxel Shuffle (3S) mitigates feature homogenization by constructing candidate child features in the channel dimension, jointly transforming and normalizing them, and then rearranging them onto a finer sparse grid with explicit feature-coordinate correspondence. Second, Entropy-Regulated Layer-wise Memory (ER-LM) retains semantic information across decoder layers by maintaining a layer-wise memory for each query through gated recurrent updates and using prediction entropy to regulate the historical information propagated to the next layer. During training, a class-weighted entropy loss Lcwe provides label-guided regularization for matched semantic queries. On SemanticKITTI, 3S increases the spatial coefficient of variation of normalized voxel-wise root-mean-square (RMS) responses from 0.41% to 4.03%, while ER-LM reduces semantic degradation and improves error correction across decoder layers. The combined FM-Occ improves mIoU from 12.16% to 12.73% with negligible memory overhead. ER-LM further improves mIoU in the BEV-based DISC framework from 16.78% to 17.17%, supporting its applicability across 3D voxel-based and BEV-based architectures.

Mixed-integer programming solvers expose hundreds or even thousands of parameters whose best setting varies per instance, yet per-instance search is prohibitively expensive. We frame this as configuration for heterogeneous Big Data optimization workloads and propose a nominate-then-adjudicate pipeline built around two design choices. First, a fixed portfolio is sampled once and reused across instances, removing per-instance configuration search at inference; on the evaluated families, the resulting shortlist retains substantial oracle headroom, making selection within it the empirical focus. Second, the fixed 20-feature representation can map structurally different instances to the same vector, preventing a feature-only selector from distinguishing them even when they prefer different configurations. The foundation model receives no task-specific weight updates, and selection is one-pass: one model call and no solver calls at inference. The study represents structural Variety through distinct formulation families and decision Value through solver-performance outcomes; the one-call, zero-solve protocol specifies the decision interface and leaves end-to-end latency unmeasured. Among comparisons exceeding a 20% practical-difference threshold, the complete adjudicator records 538 wins and 185 losses against the top nominee, while its win rate against the default rises from 84.9% to 90.2%. Against per-instance configuration baselines under a matched one-pass protocol it leads the cluster-level methods and the best-performing evaluated per-instance baseline, ISAC+.