Papers

As machine learning (ML) models are increasingly deployed in high-stakes environments, explainable AI (XAI) methods like SHAP and LIME have become essential for regulatory compliance and trust. However, the current auditing paradigm relies on an implicit "chain of trust" where third-party auditors are assumed to be trusted. Recent research demonstrates that this assumption is flawed and adversarial auditors can manipulate XAI explanations through manipulation attacks such as output shuffling or scaffolding out-of-distribution (OOD) to conceal model biases while maintaining high prediction accuracy aiming for fairwashed explanation. In this paper, we introduce a novel defense framework, ExplainGuard, that leverages a Zero-Trust architecture (ZTA) design to be incorporated within the XAI explanation supply chain and ensures the integrity of the generated explanation. This framework would help us to replace the ambiguous default assumption of "auditor is trustworthy," with a continuous "verify-then-trust" approach. Our design architecture establishes a Policy Decision Point (PDP) that enforces three distinct pillars of verification before any explanation is released to the user: (1) asset integrity via behavioral fingerprint to detect model substitution, (2) semantic validity using axiomatic consistency checks to reject mathematically impossible explanations, and (3) feature faithfulness verification utilizing a ranking stability approach with minimal computational overhead. Finally, we evaluate how ExplainGuard can effectively neutralize state- of-the-art explanation manipulation attacks while transforming the auditing process into a verifiable operation.

Knowledge-based Visual Question Answering (KB-VQA) aims to answer queries that necessitate reasoning over external knowledge sources beyond the visual content. Typically, current methods fuse multimodal features to retrieve external information, subsequently leveraging Multimodal Large Language Models (MLLMs) to derive answers from the retrieved evidence. However, these methods often struggle to capture structural associations within complex contexts to effectively filter noise. Furthermore, they frequently fail to ensure that the reasoning process remains strictly faithful to the retrieved evidence. To address these challenges, we propose SAFE-G, a Structure-Aware Faithful Evidence-guided Generation framework, which enables precise evidence localization and trustworthy reasoning. Specifically, we first employ a coarse-grained hybrid search fusing visual and textual modalities to recall candidate documents, and subsequently implement a structure-aware fine-grained graph retrieval that captures structural dependencies to filter noise and pinpoint precise evidence. Moreover, we introduce a reinforcement learning (RL) strategy with an evidence-grounded reward that assigns credit to correct answers only when the selected evidence is correct. This strict alignment constraint compels the model to anchor its response in the retrieved context, effectively enhancing its capability to locate evidence via multimodal features and perform faithful reasoning. Extensive experiments on the Encyclopedic-VQA and InfoSeek benchmarks demonstrate that SAFE-G outperforms prior methods by a margin of 8.9% and 3.5%, substantially enhancing the overall reasoning accuracy. Our source code is publicly available at: https://github.com/MINE-USTC/SAFE-G.

GUI grounding evaluations that expose UI elements as text metadata often treat high instruction-element embedding similarity as evidence of semantic grounding. Across three mobile and web benchmarks, we show that this interpretation is frequently confounded by visible-label recovery. Lexical baselines remain competitive at top-1, label-poor targets remain weak for text-only methods, and encoder top-1 hits are predictable from lexical rank, candidate-pool size, and label type. We evaluate each action as a same-screen ranking task, comparing five off-the-shelf single-vector encoders with lexical baselines. Encoders recover some lexical misses, but deployable fusion gains are much smaller than target-aware oracle gains. These findings show that embedding-based evaluations can conflate visible-label recovery with semantic GUI grounding. Embedding-based evaluations should therefore report lexical baselines, label-type stratification, and deployable-fusion diagnostics. Our released repository provides analysis scripts and detexted per-step panels: https://github.com/qijia123/lexical-coupling-release.

We study fixed-layout local verification with target-dependent local tests. Let $M$ be a random variable on $\{0,1\}^K$, and let $S$ record the test selected at each coordinate. For each $s\in\operatorname{supp}(S)$, let $F_s$ be the corresponding target fiber and set $D_{\mathrm{fib}}=\max_s\operatorname{VCdim}(F_s)$. We prove $H(M\mid S)\le \log_2\!\left(\sum_{j=0}^{D_{\mathrm{fib}}}\binom Kj\right)$. A fiber that shatters $d$ coordinates yields a weak relaxed locally decodable code with message length $d$ and block length $d+P$ over the original proof alphabet. For a uniform $K$-bit target and fixed proof alphabet, $Q$, and $\sigma$, the Goldberg–Gur–Saraogi lower bound implies that $I(M;S)\le\gamma K$, for fixed $\gamma<1$, forces $P=\Omega\!\left(K^{1+1/a}/(\log K)^{2+2/a}\right)$, where $a=\lceil Q/\sigma\rceil$. If $P\le K(\log K)^c$, then $I(M;S)\ge K-O\!\left(K^{a/(a+1)}(\log K)^{3+ac/(a+1)}\right)=K-o(K)$. Any discrete verifier state $T$ determining $S$ satisfies the same information lower bound. Bounded-randomness adaptive branches can be simulated nonadaptively by exposing their decision trees. A branch using at most $r$ random bits and $q$ adaptive proof queries yields a decoder with perfect completeness and at most $1+2^{r+1}\sum_{j<q}A^j$ queries. Under $I(M;S)\le\gamma K$, near-linear proof length requires this quantity to be $\Omega(\log K/\log\log K)$; the binary one-query case gives $P=2^{\Omega(K)}$. Applied to a global list-sound dPCP interface of Gur–Minzer–Weissenberg–Zheng, our bound shows that a fixed target-independent menu of $L$ test profiles must satisfy $\log_2L\ge K-o(K)$. Proof-dependent lists and additional target-dependent selection data must be included in the measured state.

Document classification in regulated industries is constrained by data residency, limited cold-start labels, scarce review capacity, and costly model-governance procedures. We present HIRA, a training-free, on-premises retrieval-augmented cascade for document classification in regulated deployments that combines BM25 over OCR text, dense text embeddings, and image-level representations through validation-calibrated weighted reciprocal-rank fusion. Confident documents are classified directly by retrieval; uncertain or visually confusable documents are passed to a locally hosted LLM verifier, which receives the OCR text, retrieved exemplars, label descriptions, and confusion-specific terms. When the verifier remains uncertain, the document is sent to human review. Each correction is stored as a margin-weighted retrieval exemplar and updates a Dirichlet-smoothed confusion graph, letting the system improve without updating model weights. On a private 80-class trade-finance corpus, HIRA processes the full 30,233-document production stream while requesting human correction for only 1,945 documents (6.4%), improving Macro-F1 from 0.6218 to 0.8548. On the corrected Tobacco-3482 benchmark, HIRA reaches 0.9423 Macro-F1 with a locally hosted DeepSeek-R1-Distill-Qwen-32B verifier, 17.4 percentage points above the zero-shot LLM baseline, while invoking the verifier for only about 40% of documents and reducing LLM calls by approximately 60%. With 518 human corrections (24.8% of the pool), HIRA matches the fully labelled pool oracle, in which all 2,086 pool documents are indexed with their ground-truth labels. These results show that selective human feedback and retrieval-memory adaptation can be a practical alternative to repeated model retraining for long-tail document classification in regulated deployments.

Metric distortion has primarily been studied for social choice functions, which select a single winner from ordinal preferences. We extend this framework to social welfare functions, which output a ranking of $m$ candidates. We associate each voter $v$ with a monotone weight vector $\mathbf{w}_v = (w_{v1},\ldots,w_{vm})$, specifying the importance of the $i$-th position for voter $v$, and define the cost of a ranking as the position-weighted sum of their distances to the ranked candidates. This model generalizes both single-winner voting and committee selection. We consider three information regimes. First, we study the setting where the positional weight vectors are known. A natural approach recursively applies a single-winner rule with distortion $\beta$ to construct the ranking one position at a time. We show that this yields distortion at most $3\beta$ in general. This gap is not merely an artifact of the analysis: we show that no analysis based solely on per-round guarantees can certify a bound better than $2\beta$. By exploiting structural properties specific to Fractional Veto of Kizilkaya and Kempe, we show that its recursive extension achieves the optimal distortion of 3. Second, when all voters share the same unknown weight vector, recursively applying any social choice function with distortion $\beta$ achieves distortion at most $1+(\beta-1)\text{range}(\mathbf{w})$, where $\text{range}(\mathbf{w})=(w_1-w_m)/w_1$ denotes the normalized range of the common weight vector $\mathbf{w}$. Finally, we study unknown heterogeneous weights. Without further assumptions, every rule has unbounded distortion. We therefore consider two natural normalizations: unit-sum, where each voter distributes one unit of value across the ranking, and unit-top, where every voter assigns unit value to the first position. Under both models, we show that the optimal distortion is $\Theta(m)$.

We study the linear recurrence order of the Potts model partition function on thermal 2D grid graphs. By restricting the transfer matrix (TM) to the real coupling axis, the physical operator maintains diagonalizability under the Spectral Theorem. We show that this thermal regularization allows the Krylov subspace to saturate the unconstrained planar state capacity, locking the recurrence order to the Dyck path up to reversal (OEIS A007123) for $q \ge 4$. Furthermore, the recurrence order collapses to height-restricted Dyck paths up to reversal (OEIS A001998) for $q=3$ due to finite-index Jones-Wenzl projections, and to the zero-magnetization conservation sector (OEIS A001405) for $q=2$. This framework bridges graph-theoretic combinatorics with the representation theory of physical loop gas models.

General image fusion aims to integrate complementary information from multiple source images, but existing methods often rely on task-specific models and struggle to maintain robust performance under diverse degradation conditions. In this paper, we propose UniDiffFusion, a unified diffusion framework for multi-task and degradation-robust image fusion. UniDiffFusion leverages the strong generative prior of a pretrained diffusion model to establish a shared fusion backbone across heterogeneous fusion tasks, while introducing task- and degradation-aware conditional adaptation to accommodate their distinct information-selection requirements. Specifically, we employ task prompt modulation to progressively adapt the shared diffusion representations to different fusion objectives, and develop a degradation prompt router to dynamically retrieve degradation-aware priors and restore corrupted source features before fusion. Furthermore, an application prompt bank is introduced to incorporate task-oriented semantic guidance for downstream applications, such as object detection and semantic segmentation, without altering the shared fusion and restoration pathways. The proposed framework is trained in a progressive manner to decouple fusion learning, degradation-aware restoration, and application-specific adaptation, thereby reducing interference among heterogeneous objectives. Extensive experiments on visible-infrared, multi-exposure, and multi-focus image fusion demonstrate that UniDiffFusion achieves superior fusion quality and robustness under both clean and degraded conditions. Moreover, UniDiffFusion consistently improves downstream detection and semantic segmentation performance, demonstrating its effectiveness as a unified diffusion framework for both perceptual fusion and task-oriented vision.

Few-step text-to-image models increasingly replace slower generators, yet acceleration can silently change distributions over unspecified attributes even when individual outputs remain plausible and aligned. We call these distributions semantic defaults and their change under replacement semantic default shift. Existing quality, preference, and diversity evaluations do not test whether a replacement preserves its reference model's semantic defaults. We introduce DefaultShift, a paired audit that labels repeated samples with closed semantic vocabularies, measures probability-mass movement, and separates interpretable ranking from confirmatory cross-fit inference. Across 14 reference and replacement pairs, adjusted color discrepancies range from 0.054 to 0.303 with recipe-specific directions. A 1,000-image human audit reproduces the ordering. We further introduce DefaultShift-Select, an offline calibration method that reduces human-measured shift by 10.3 percent to 35.1 percent across Turbo, DMD2, and FLUX without material quality loss. Under balanced evaluation, selected data recover 4.3 accuracy points and 7.5 worst-group points over uncalibrated replacement data. DefaultShift makes semantic preservation under acceleration measurable and actionable.

Dimensionality-reduction (DR) methods are routinely judged by how well each point's k nearest neighbors survive the 2-D embedding (recall@k, trustworthiness, continuity). We argue this family is a biased measure of distance fidelity: its per-point variable radius and hard inclusion threshold favor neighbor-graph methods (t-SNE, UMAP) and penalize methods that preserve absolute distances. We instead score DR fidelity with a fixed-radius distance-band Shepard rho: the Spearman correlation between high-D and 2-D pairwise distances, restricted to cumulative distance bands so that near and global structure are reported separately, with every point judged on the same absolute radius. On synthetic datasets with known ground-truth geometry (non-uniform density, dense clusters, a closed-loop transition, off-subspace outliers, imbalanced two-population data) at realistic noise (SNR=1, D=768, N=1000), we benchmark eight methods – PCA, Isomap, t-SNE, UMAP, PyMDE, PCC, DREAMS, and the closed-source toorPIA – and show that (i) high global Shepard rho can coexist with a ~93x collapse of within-cluster scale, invisible to rank-based scores but obvious in a value-based over-compression metric; (ii) recall@k and the fixed-radius band disagree systematically, in the direction the bias predicts; (iii) a membership-restricted Shepard rho resolves single-point and minority-population questions that many-pair statistics cannot – questions on which even DREAMS, a recent local-plus-global hybrid, fails silently. A supplementary out-of-sample (addplot) test asks whether a never-seen anomaly lands outside the normal region and whether its direction identifies its source. All metrics are computed exactly on all pairwise distances, independently of any method's internals, and every number is reproducible offline: the closed-source method's output coordinates (not its algorithm) are committed to the artifact.

Autonomous excavation requires an intelligent control system that can convert spatial work intent into coordinated bucket motion under contact-rich soil interaction. This paper presents a target-conditioned intelligent control framework for autonomous excavation in a physics-based deformable-soil simulation workflow. An image-aligned target mask serves as a visual spatial command for the desired digging region, while a mask-conditioned Action Chunking Transformer maps multi-view RGB observations, proprioception, and the target mask to temporally extended joystick commands. To reduce target-ignoring behavior, demonstrations are organized with paired-condition supervision, where the same or closely matched scene is demonstrated with different target masks and corresponding action chunks. The framework is evaluated through both a diagnostic manipulation task and an excavation simulation benchmark with single-scoop and sequential pile-clearing protocols. In manipulation, target success is 4% for no-condition ACT, 63% for non-paired mask-conditioned ACT, and 96% for paired-condition mask-conditioned ACT. In sequential pile clearing, paired-condition mask-conditioned ACT removes 76.8% of the pile versus 27.4% and 15.7% for the two baselines, with 91.0% human-normalized efficiency. The results show that visual target conditioning, paired demonstration structure, and action-chunk control form a practical cyber-physical simulation pipeline for excavator automation.

Recent advances in image editing allow impressive manipulation of objects, existing methods still struggle to handle spatial movement in complex scenes, such as objects span different depth layers or are partially occluded. Most image editing methods focus solely on prior information from 2D datasets, emphasizing planar features while lacking support for spatial structures. Even approaches that incorporate explicit positional information fail to capture true 3D spatial relationships, thus limiting accurate object movement in complex scenes. In this paper, we present SpatialDiff, a method that effectively captures 3D spatial structures, enabling precise and consistent object movements in complex scenes. Our core innovations are twofold: (1) Implicit 3D Spatial Modeling, which introduces 3D prior knowledge and enables the model to internally build a comprehensive understanding of the three-dimensional spatial structure; and (2) Global Spatial Supervision, which constrains the latent spatial features to enable the model to perceive changes in object spatial positions caused by editing operations. Experimental results demonstrate that our method significantly improves the accuracy and fidelity of spatial movement in complex scenes.

Large Language Models (LLMs) are increasingly deployed in real-world applications, yet they remain vulnerable to generating harmful content. From adversarial jailbreaks that bypass safety filters to implicit hate that evades detection, the range of risks these models pose continues to grow. While both specialized content moderators and general-purpose LLMs are being used as safety layers, the question of which model is best suited for which type of harmful content remains unanswered. We present the most comprehensive evaluation of LLM safety capabilities to date, systematically testing 53 models across 11 datasets that we organize into four distinct categories. Our evaluation under both prompt-only and prompt-response settings uncovers critical blind spots: large frontier models that lead on one category fall significantly behind smaller, specialized alternatives on others, and real-world conversational safety remains largely unsolved across all model families. These findings challenge the assumption that scale alone ensures safety, and provide the community with a structured framework for informed model selection.

Semantic communication has emerged as a promising paradigm for next-generation wireless systems by transmitting high-level semantic features rather than raw bits. However, collaborative devices and multimodal transmission increase privacy risks because sensitive information may leak through inter-device semantic fusion and cross-modal representations. To address this issue, we propose a privacy-preserving semantic communication framework for wireless edge networks. Leveraging a vision-language model (VLM), the framework extracts textual semantics from images and identifies privacy-sensitive entities using a privacy database maintained only at the edge server. Before image transmission, each device removes the identified private regions while preserving useful semantic content. The server then reconstructs the removed regions from the received masked images using textual embeddings and VLM-based semantic priors. To protect textual information, we design an encrypted semantic-channel transceiver using physical-layer keys generated from reciprocal wireless channels, without pre-shared keys. We also introduce a semantic information bottleneck to suppress redundant information across multiple devices. The framework is evaluated against a strong model-aware adversary that can intercept wireless transmissions and access edge-device model parameters but not server-side data. Simulation results show that the proposed method reduces privacy leakage by more than 50% compared with a semantic communication scheme without privacy protection, while the authorized server achieves a 48% improvement in perceptual reconstruction quality over the adversary. The estimated mutual information between transmitted representations approaches 0 bit, indicating effective suppression of cross-device semantic redundancy.

Accurate modeling of sea ice concentration (SIC) evolution is essential for polar climate assessment and short?range sea ice prediction. Numerical and data-driven approaches constitute major foundations for SIC modeling, but the former often require complex parameterizations and substantial compu?tation, whereas the latter rarely encode physical dependencies explicitly. This study presents the Physics-Informed Hybrid Ice Model (PIHIM), a differentiable data-driven hybrid ice model for daily SIC evolution that organizes its network structure according to the physical dependencies encoded in the sea ice continuity equation and explicitly accounts for dynamical transport, ther?modynamically driven areal growth and loss, and unresolved local processes. PIHIM preserves the representation capacity of deep learning while providing a process-decomposed formulation of ice displacement, freeze-melt areal change, and local error closure. Two evaluation settings are adopted: reanalysis-forced simulation examines SIC evolution stability under reanalysis forcing, and forecast-forced prediction assesses short-range performance un?der forecast-forced conditions, with reanalysis and observational SIC serving as verification references. Results indicate enhanced ice-edge preservation and error-growth control in reanalysis?forced simulation, while PIHIM retains measurable short-range prediction skill under forecast-forced conditions. Our code will be made publicly available after the paper is accepted.

Both capability and safety benchmarks rest upon the assumption that the behavior of language models undergoing a test is informative about their behavior in deployment. This assumption can fail, should models infer that they are being evaluated and condition their response on such context. This hypothesis, termed “evaluation awareness”, has been observed in frontier and open-weight language models alike. We provide a systematic study of this phenomenon, by probing for it across six language models (from four families and three sizes) and three metrics. More precisely, we examine whether (i) being under evaluation is linearly represented within the models' activations space, (ii) it is verbalized in their output tokens (as scored by an LLM-as-judge), and (iii) steering causally affects their behavior. For the open-checkpoint Olmo models, we further test these measures at every training stage. In doing so, we report that evaluation awareness is linearly decodable from the residual streams of every model (best AUROC $\geq 0.7$). By contrast, these representations align only in part with verbalization: their correlations and mutual information are nonzero in some settings, yet vary substantially across models, layers, and readout choices. Nevertheless, steering along probe-derived directions can shift the verbalization scores. Finally, a comparison across the Olmo checkpoints reveals that evaluation awareness is already present within base models, becomes amplified throughout the stages of supervised fine-tuning, and remains stable thereafter—unlike the effects of steering, that grow more pronounced at every successive training stage. These results show the need for evaluations to account for the disjunction between what models represent internally, what they verbalize, and their steering.

Event-based vision has emerged as a promising paradigm for energy-aware artificial intelligence (AI), offering sparse, low-latency visual signals that reduce redundant data processing and support sustainable edge computing. However, the asynchronous and noise-prone nature of event streams creates challenges for conventional deep learning models, which are often too computationally intensive for low-power embedded platforms. This work presents a compact and configurable event-driven autoencoder that efficiently compresses neuromorphic data while preserving essential spatiotemporal structure for downstream inference. The architecture integrates lightweight convolutional encoding with robust performance under adaptive event thresholding and a minimal classifier head, enabling substantial reductions in computational cost without degrading recognition fidelity. Extensive evaluations on the Smart Event Face Dataset (SEFD) and Event-Based Crossing Dataset (EBCD) show that the proposed framework achieves competitive or superior accuracy compared to YOLOv9 while requiring up to 35.6$\times$ fewer parameters. To assess real-world sustainability, the model is deployed on resource-constrained hardware: a Raspberry Pi 4B and a NVIDIA Jetson Nano. On NVIDIA Jetson Nano, it delivers real-time throughput of 44.8 FPS. On a Raspberry Pi 4B CPU, the 50% autoencoder classifier consumes 16.19 J for the evaluated inference workload, corresponding to approximately 726.3$\times$ lower energy consumption than YOLOv9 under the same evaluation protocol. These results demonstrate the potential of compact event-driven models to advance environmentally conscious, low-power AI systems for high-speed perception in autonomous, mobile, and embedded computing environments.

Data-driven methods for earthquake and tsunami early warning rely on large ensembles of rupture scenarios and their resulting waveforms, but generating such datasets with repeated high-fidelity seismic and tsunami simulations is prohibitively expensive. We exploit the linear time-invariant structure of both dynamics to precompute elastic Green's functions and acoustic-gravity adjoint responses, reducing the source-to-waveform map to two consecutive convolution operators. We evaluate these convolutions with a distributed, FFT-accelerated GPU pipeline that partitions the large seafloor grid across GPUs and directly generates the final observation waveforms. We demonstrate the scalability of this pipeline for the Cascadia Subduction Zone with 963 subfaults, 2,416,530 seafloor grid points, 64 observation locations, and 256 timesteps, requiring 9.45 TiB of aggregate GPU memory. On 64 GB200 GPUs within one NVL72 domain, the pipeline generates waveforms in 24 ms per rupture once the response operators are resident, enabling large rupture ensembles to be evaluated within minutes.

Vision-language models (VLMs) fail many detail-centric questions for a concrete reason: the answer is visible in the image, yet lost after the image is compressed into a low-resolution global view. Allocating more visual tokens to every query improves some OCR and document cases, but it spends computation indiscriminately and can disturb tasks that rely on global context. We propose GapSight, a framework for learning visual re-reading: a VLM first takes a global glance, then selectively returns to a free-form region when the question calls for local evidence. The supervision comes from the target model's own failure signal. Offline, we compare answer loss or multiple-choice option margin under a global-only view and candidate crop-augmented views; crops that improve the target answer become model-specific review labels. A lightweight free-form crop router distills these labels into a one-shot inference policy that predicts whether to review, expected utility, and a continuous crop box from the global state. Across LLaVA-1.5-7B, InternVL2.5-8B, and Qwen2-VL-2B-Instruct, GapSight improves the Base no-zoom baseline on six benchmarks spanning OCR, documents, charts, infographics, VStarBench, and MME-RealWorld-Lite. On InternVL2.5-8B, GapSight raises the six-benchmark average from 52.25 to 64.29, above CropVLM (57.16), ViCrop (55.84), and ZoomRefine (54.43). Mechanism analyses show that the router rescues concrete wrong answers, adapts its action rate by task, and forms a favorable token-performance profile. These results position loss-gap supervision as a practical route to teaching VLMs when and where to look again.

Large collections of street-view imagery provide rich visual information about urban environments, but extracting fine-grained geographic information from such data remains challenging. In particular, fine-grained regional geolocalization is challenging because nearby areas often share coarse geographic cues. We study regional geolocalization within a metropolitan area and ask whether pretrained CLIP features are sufficient for regional discrimination, and what visual information supports performance after adaptation. Using 9,085 street-view images from eight Greater Los Angeles regions, we compare zero-shot CLIP, frozen-encoder readouts, partial encoder updating, Low-Rank Adaptation (LoRA), and full fine-tuning. Frozen readouts remain near the 39.03% zero-shot accuracy, whereas encoder adaptation achieves 75.94-82.10%. Full fine-tuning also reduces the mean distance to the predicted region center from 12.30 km to 3.86 km. We probe these gains through semantic cue removal, appearance reduction using edge maps and blur, and scene-configuration disruption using patch scrambling. Adapted models achieve higher edge and blur accuracy and switch 42.92-45.56% of predictions after scrambling, compared with 10.79-14.60% for frozen methods. However, adaptation does not improve the fraction of performance retained after appearance reduction, while vegetation and sky remain influential. A Caltech101 control further shows that scrambling sensitivity is not unique to geolocalization. Overall, encoder adaptation substantially improves nearby-region discrimination and is associated with greater sensitivity to intact scene configuration, without evidence that coarse structure alone becomes sufficient for prediction. These conclusions concern viewpoint variation near known locations rather than geographically disjoint generalization.

This paper proposes modified polarization-adjusted convolutional (MPAC) codes and their hybrid decoding that achieves an improved performance-complexity tradeoff. For MPAC codes, only a subset of the information bits undergo the convolutional transform. The output is then combined with the remaining information bits for the inner polar transform. Correspondingly, the convolutionally transformed bits are recovered by Fano decoding, while the remaining information bits are recovered by the successive cancellation (SC) decoding, constituting the hybrid Fano-successive cancellation (HFSC) decoding. The MPAC codes are further designed by the coset-wise analysis that characterizes the number of minimum weight codewords (MWCs). It is discovered that a partially convolutional transform can improve the codeword through utilizing the row combinations of the frozen set efficiently. This property enables the MPAC codes to outperform their prototype polarization-adjusted convolutional (PAC) codes and cyclic redundancy check (CRC)-polar codes. Furthermore, MPAC codes can be optimized by reducing the number of MWCs. Our numerical results demonstrate that, with a similar decoding complexity budget, the MPAC codes offer competent decoding performance when compared with PAC codes using Fano decoding and CRC-polar codes using SC list (SCL) decoding.

Deep-learning efforts have increasingly shifted toward foundation model approaches. In experimental physics, this allows models and learned representations to be reused beyond the experiments in which they were developed. This work evaluates the reusability of representations across experiments and detector systems using probes on frozen encoders. These probes reveal task-relevant structure before downstream adaptation, complementing fine-tuning. Together with random-weight controls, they distinguish contributions from architecture and encoder training that downstream performance alone cannot resolve. Time projection chamber (TPC) data provide a useful testbed because events from TPC systems can be represented as variable-length sparse tensors, while detector geometries, event topologies, and scientific tasks can differ substantially. We investigate whether fixed-dimensional TPC event representations can be reused across classification tasks, experiments, and detector systems. Sparse ResNet and PointNet-style encoders produce 512-dimensional embeddings for four datasets from the GADGET II TPC and AT-TPC. Randomly initialized encoders isolate the contribution from architecture before supervised training. We then train each encoder on a classification task, freeze its parameters, and train a linear or nonlinear probe for each downstream task. We find that this architecture-induced structure remains useful across experiments and detector systems. The randomly initialized PointNet-style representation is highly informative on several tasks. The two architectures organize their embedding spaces differently, but neither exhibits a large, systematic loss of utility cross-detector. These results show that architecture is a major source of task-relevant structure in TPC embeddings and should be treated explicitly when assessing representation learning and developing reusable detector models.

Long-horizon agents require more than retrieval over stored conversations: they must preserve source evidence, revise outdated knowledge, and distinguish relevant memories from sufficient support for an answer. We present ECHO (Embodied Context & History Orchestration), an auditable dual-path memory system. A fast path captures immutable events and serves multi-route recall, while asynchronous workers consolidate experience into revisable state and rebuild derived indexes. Source provenance, temporal authority, and answer realization are explicit boundaries. The service supports original-source packets with evidence-use checks and disables model-based reranking by default. For the latest reported accuracy, the authors confirm mixed retrieved inputs comprising original dialogue and memory summaries, answered by Qwen3-8B with NF4 quantization. In the latest author-supplied five-system aggregate evaluation, ECHO reports 56.27% answer accuracy, 98.04% source Hit@10, 93.58% macro source Recall@5, and 13.92% hallucination flags, with 19.54 ms mean retrieval latency. Its reported accuracy exceeds Mem0 by 4.53 percentage points in this snapshot. The results are descriptive aggregates; matched question-level evaluation and generation latency remain to be established. We separate these observations from earlier published experiments and engineering regressions.

Despite the success of data pruning (DP) in reducing training data sizes and improving downstream model performance in classification and segmentation tasks, its potential in remote sensing change detection remains unexplored. For the first time, we benchmark six representative DP methods across building- and forest-change datasets, CNN- and transformer-based models, and three pruning budgets, and show that existing baselines yield no reliable advantage over random selection. Notably, even the strongest evaluated baseline, Feature Diversity, is matched or exceeded by $\sim$33% of randomly sampled subsets. To understand the underlying mechanism, we conduct a systematic regression study over 540 randomly sampled data subsets, characterizing each with four descriptors covering label statistics, image diversity, and feature-space geometry. Random Forest models show that change distribution fidelity is the most prominent factor in determining the quality of change detection data subsets, a property absent from the existing pruning literature. Our analyses further show that pixel-wise image diversity and label-feature consistency are secondary factors. We translate these findings into Fidelity-Diversity-Consistency (FDC), a simple two-stage pruning method that shows consistent improvements over existing baselines across change detection benchmarks and backbones, especially at lower pruning ratios. Code is available at \href{https://github.com/ddydyd32/fidelity-diversity-consistency}{https://github.com/ddydyd32/fidelity-diversity-consistency}.

It has long been known that well-trained neural networks can be compressed very strongly without affecting their performance, an important phenomenon that remains poorly understood. We prove a uniform compressibility theorem for deep multilayer perceptrons with analytic activations. For a deep, wide fixed teacher network, there exists a narrow (same depth) network that approximately represents the same function as the original. The reachable compressed width is strikingly independent of the original width, but is $O((\log(1/\varepsilon))^{d_{in}})$, where $\varepsilon$ is the error budget and $d_{in}$ is the effective input dimension. Our construction involves a novel derivative-matching technique which is aware of the low-dimensional input, and a layer-wise reweighting that preserves the input-output mapping.

Parametric retrieval-augmented generation (PRAG) injects retrieved evidence into a large language model (LLM) through passage-specific LoRA adapters, reducing reliance on long in-context prompts. When multiple passages are retrieved for the same query, however, evidence-level fusion becomes a bottleneck: equal-weight merging can amplify weak or conflicting evidence, and translating retrieval signals into fusion weights often requires fragile global tuning. We propose FCPRAG, a fusion-controlled parametric RAG framework that adds a lightweight controller for retrieval-conditioned, sample-level adapter fusion. The controller predicts per-passage fusion scores together with sample-level calibration signals, including a mixing gate and an adaptive temperature, enabling fusion that stays selective under informative retrieval signals and conservative under uncertainty. FCPRAG is trained with merge-aware supervision derived from each adapter's marginal contribution within a multi-adapter merge, using training data only. We further show that a single dataset-level temperature is suboptimal under heteroscedastic retrieval uncertainty, motivating sample-level adaptation. Experiments on HotpotQA, 2WikiMultiHopQA, PopQA, and ComplexWebQuestions (CWQ) across three LLM backbones show that FCPRAG consistently improves F1 over standard RAG and parametric RAG baselines, with gains of up to 4.65% on 2WikiMultiHopQA and 7.55% on CWQ, while also reducing tuning cost and improving robustness under retrieval perturbations.

Verifier-guided text-to-image systems increasingly use test-time search to select, refine, or stop among multiple candidates, yet release thresholds are often calibrated on individual images. This creates a candidate-to-policy calibration mismatch: search changes both which prompts receive an output and which candidate is released, so candidate-level risk control need not imply control of released-output risk. We formalize this estimand shift through prompt reweighting and within-prompt selection, and introduce SHIP, Selection-aware Held-out calibration of Inference Policies. SHIP runs or replays the complete deployed policy on held-out prompts, evaluates the image it actually releases using an independent target judge, and selects the most permissive threshold whose risk upper bound satisfies a prescribed budget. For replayable policies with a prespecified threshold grid, simultaneous confidence control provides finite-sample validity. Experiments across fixed, sequential, and adaptive T2I inference procedures show that policy-level calibration recovers lower-risk operating points while exposing policy-dependent tradeoffs among risk, coverage, and compute. On GenEval2 with FLUX at N=16, a pooled-candidate threshold yields released risk 0.310, whereas SHIP reduces it to 0.162. Across 200 cached-stream splits, the fixed-grid certificate has no target crossing. Reliable inference-time scaling therefore requires calibrating the output distribution induced by the complete deployed policy.

LLM-based coding agents generate complete software systems from high-level descriptions, yet little is known about how the format of architecture specifications affects the quality of generated code or whether this effect depends on model capability. We present a controlled experiment comparing five informationally equivalent specification formats (informal prose, Mermaid diagrams with constraints and ADRs, OpenAPI, C4/Structurizr DSL, and TypeScript interface contracts with ArchUnit-style rules) across six models from three vendor families (Anthropic Claude, OpenAI GPT, Google Gemini). Across 90 multi-turn agent trials, specification format shows a strong format x model interaction. On the strongest models (Sonnet 4.6, GPT-5), format barely matters (quality spread 0.17-0.92). On weaker models, format produces spreads of 0.83-2.42 points, with code-proximate formats (OpenAPI, TypeScript contracts) recovering most of the capability gap. Mid-tier models can consume more tokens than frontier models for worse output when they enter compilation debugging loops that stronger models avoid. Self-validation rates collapse from 100% (Sonnet) to 0% (Gemini Flash) across the capability spectrum. TypeScript contracts triple API route coverage for the weakest model (33% to 100%). Structured architecture specifications serve as a capability equalizer, with value inversely proportional to model strength and the largest returns for cost-optimized deployments.

Language grounding increasingly involves non-text observations whose structure is not naturally expressed as words or objects. We study this problem for physiological time series in non-contact blood pressure (BP) estimation: remote photoplethysmography (rPPG) provides measured evidence about bodily state, but numerical pipelines expose little semantic structure about why a window is reliable or how its cues should be fused. We introduce observation-matched physiological semantic grounding, where language-derived priors must be constructed from the same rPPG observation, remain bounded by an auditable prior contract, and avoid BP-label or identity leakage. CALM-BP does not treat language as new physiological evidence; instead, it verbalizes rPPG descriptors into a controlled semantic interface while rPPG remains the primary haemodynamic evidence source. FlowBP-Set pairs forehead observations, synchronized BP labels, and structured physiological prompts from 81 participants. Main BP results, direct cross-dataset evaluation, language-realization ablation, and observation-mismatch controls test whether language helps because it organizes the current physiological observation rather than because it is arbitrary auxiliary text. The FlowBP-Set dataset contains sensitive facial video and physiological recordings and is therefore not publicly available due to privacy and ethical restrictions. Data access may be considered upon reasonable request and subject to applicable ethical and institutional approval.

Learned joint source-channel coding (JSCC) enables robust wireless image transmission by jointly optimizing the transmitter and receiver over differentiable channel models. For bandwidth-limited and time-varying visual links, a single model should support user-adjustable transmission rate and adapt to changing wireless channel conditions, while also dynamically allocating resources according to spatial content. Existing content-adaptive or dynamic allocation schemes often rely on entropy coding, context/probability prediction, explicit rate maps or masks, or auxiliary allocation networks, complicating the encoder-decoder pipeline and increasing side-information overhead. We propose TS-JSCC, a single-model adaptive JSCC framework with tail-structured sparsification. First, an L1-based tail-structured sparsification objective encourages each token to retain an active feature-channel prefix while suppressing trailing ones. This enables content-adaptive feature-channel allocation with compact side information through active-prefix transmission. Second, lightweight stage-wise neural regulating modules use a normalized sparsity-control coefficient and the channel signal-to-noise ratio (SNR) to rescale intermediate features for single-model transmission rate and SNR adaptation. Experiments on CIFAR-10, Kodak, and CLIC2021 under additive white Gaussian noise (AWGN) and Rayleigh fading show that TS-JSCC achieves strong rate-distortion performance against the latest learned-JSCC baselines and remains competitive with the considered idealized separation baselines, while retaining a simple one-shot encoder-decoder without extra structures or computations.