In this paper, a worst-user outage probability minimization problem in active reconfigurable intelligent surface (RIS)-aided systems is addressed. Since the resulting problem involves a discontinuous objective function over an unknown channel distribution, a sequence of reformulations is applied by integrating a probability-simplex user weighting, a softplus surrogate for differentiability, and entropic tail-risk aggregation for rare but severe outage events. Based on the reformulated problem, a data-driven stochastic optimization technique that relies on historical channel samples is adopted to jointly optimize the beamformer and reflection coefficients with projected adaptive moment estimation. Simulation results show advantages of the proposed technique for reliability-oriented active RIS transmission design.
Junyan Liu, Haipeng Luo, Lillian J. Ratliff, Zihan Zhang
原文摘要
We study online learning in two-player uninformed Markov games, where the opponent may follow an arbitrary, unknown policy in each episode and the learner does not observe the opponent's actions. [TWYS21] showed that achieving sublinear external regret requires an exponential dependence on the episode length $H$. Motivated by this limitation, they studied the weaker notion of Nash-value regret and obtained an $\widetilde{O}(K^{2/3})$ regret bound, where $K$ is the number of episodes. More recently, [LLZR26] introduced a stronger notion called empirical Nash-value regret and achieved an $\widetilde{O}(\sqrt{K})$ bound when the opponent's non-stationarity is moderate, including the case where the opponent follows a fixed policy. However, whether the $\sqrt{K}$ rate is achievable in the worst case remained open. In this paper, we resolve this open question by developing an algorithm that achieves $\widetilde{O}(H^{3/2}S\sqrt{AK})$ empirical Nash-value regret against arbitrary opponent policies, where $S$ is the number of states and $A$ is the number of learner's actions. Specifically, our algorithm combines global and local planning by maintaining a collection of candidate predictions of future losses, each associated with a policy. Global planning learns which candidates to favor across episodes, while local planning combines their policies differently at each state to make decisions. A key ingredient is to coordinate these two components to control the discrepancy between local loss estimates and the total loss of the executed policy, using only the learner's observed trajectories.
A causal video generator produces two key--value (KV) states for every segment: the KV it uses while denoising that segment, and the KV it writes into the cache for later segments to read. In a pretrained generator the two already differ substantially, later segments attend to them differently, and replacing the written KV with the generation KV degrades the continuation. We propose dKV Forcing (dKV denotes differentiable KV) to learn a dedicated memory representation from its future use: a writer-specific low-rank residual on the shared backbone separates memory writing from generation and is trained through the future reads of what it writes. Learning a write from its future use needs gradients through the cache, which the mutable cache in common use cannot provide. dKV Forcing replaces the training-time cache by persistent differentiable pages with local rematerialization and, in one temporal traversal, computes exactly the gradient that a two-pass reconstruction of the same objective computes. One traversal computes the full memory gradient at $7\%$ generator-update overhead over detached-write training, with $28\%$ less resident memory and $20.8\%$ shorter training cycles than the two-pass reconstruction. Over Self Gradient Forcing (SGF), the same objective trained with the same traversal, writer adaptation raises VBench-Long aesthetic quality by $6.6$ points at 60\,s and $8.5$ points at 240\,s with the highest subject and background consistency, and rollout dropout of the residual keeps most of this gain while restoring the motion of training without write gradients.
Ami Tavory, Yaniv Sheena, Shaked Bar, Yair Zagdansky
原文摘要
Risk screening and recommendation systems pass or block accounts that generate both value and loss. Under a blocked-value budget, a cap on expected value forgone through blocking, the optimal ranking orders accounts by expected value per unit expected loss, a ratio of conditional expectations. This rule can conflict with accepted harm or net-harm criteria, blocking accounts that another criterion would pass at the same budget. We call these cases account-level objections and define group complaint disparity as the difference between groups' normalized realized objections. Weighted binary cross-entropy (W-BCE) implements the value-per-loss order with one score. Objection penalties modify the ranking, while group multipliers scale those penalties differently by group, tracing tradeoffs between loss capture and complaint disparity. This preserves the value-budgeted ranking structure and makes the cost of reducing contestable blocks explicit. Across public recommendation, insurance, credit, and mortgage datasets, we compare undamped W-BCE with ordinary BCE using yield Gini, an area summary of loss capture versus blocked value. Logarithmic weight damping is evaluated separately. Held-out evaluations measure capture-disparity tradeoffs. Validation-selected group-aware complaint penalties improve complaint closure in some cohorts. A corrected comparison normalizes and enlarges candidate policy banks using validation-only price sweeps and retains all ten applicable sources. The source-level results do not establish a statistically significant average advantage over group-blind pricing at the fixed minimum retained loss capture. We also compare with blocked-count, blocked-value, and protected-loss parity postprocessors, which do not directly constrain realized objections, and measure objections in candidate-count and loss units.
Robotic manipulation requires accurate metric depth, yet commodity RGB-D cameras often fail on non-Lambertian objects such as transparent, reflective, and translucent materials, producing holes, structured noise, and biased depth. Most restoration methods produce only a single depth map, treating noisy and reliable regions uniformly for downstream tasks while ignoring regional uncertainty. We present UA-Depth, an uncertainty-aware depth restoration framework that iteratively co-refines depth and pixel-wise uncertainty, and uses uncertainty-gated features to suppress unreliable cues for stable updates. To strengthen zero-shot behavior, UA-Depth leverages a frozen depth foundation model as a prior to initialize missing or unreliable depth, and applies two-stage MAD-weighted log-depth scale normalization to reduce outlier bias in relative-to-metric alignment. We introduce an eight-dataset zero-shot benchmark across diverse scenes, materials, and sensors. Across these benchmarks, UA-Depth achieves the best mean rank among representative baselines, with particularly strong gains on datasets dominated by transparent and reflective objects. Beyond depth metrics, we demonstrate downstream improvements in category-level 6D pose estimation and 6D grasp synthesis by exploiting the predicted uncertainty. \emph{Code, models, and data for training and evaluation will be released upon publication.}
Edge user/service scheduling governs where and how AI services are executed under limited communication and computing resources. Existing edge scheduling frameworks typically assume numerically specified service requirements, such as latency bounds and energy budgets. However, users often express ambiguous and context-dependent expectations in natural language, creating a gap between user intent and scheduling decisions. We propose semantic contract-guided edge potential scheduling (STEPS), a natural language-driven framework that introduces semantic contracts as executable interfaces between user-side semantics and edge-side decisions. A large language model (LLM)-assisted semantic parser extracts service requirements and raw confidence scores from natural language requests. Confidence calibration and interpretable mapping rules convert these outputs into preference weights, fulfillment bounds, and semantic uncertainty. STEPS formulates intended per-slot edge scheduling as an exact potential game for execution-node selection, computing-resource provisioning, and bandwidth allocation, followed by capacity-feasible resource realization. Fulfillment-driven feedback signals enable adaptive scheduling under evolving service and network conditions. We establish the existence of pure-strategy Nash equilibrium, prove finite-step convergence of uncapped strict best-response dynamics, and characterize the boundedness of projected adaptive-control parameters. Experiments under synthetic and Melbourne-based hybrid settings demonstrate improved contract fulfillment and reduced service loss relative to several adapted benchmarks. Post-drift evaluations further show that feedback adaptation improves fulfillment and admission coverage under the tested conditions.
Temporal Knowledge Graph Question Answering (TKGQA) requires language model agents to retrieve evidence, select relevant facts, and perform multi-step temporal reasoning. However, the trajectory-level outcome signal in GRPO does not explicitly distinguish the quality of intermediate decisions: downstream failures can penalize useful evidence acquisition, while identical rewards in all-failed groups eliminate the relative outcome signal despite differences in intermediate progress. We propose Credit Redistribution with Evidence-Driven Opportunities (CREDO), which uses intermediate evidence to recover local supervision, including in all-failed groups, and determine where outcome supervision acts. CREDO assigns positive stage-level credit through comparisons among rollouts of the same question and uses retained answer overlap as an opportunity proxy to shift part of the outcome-policy weight toward reasoning and answering. The framework preserves terminal rewards and trajectory-level advantages, while redistribution conserves total outcome-policy weight per trajectory. Following the public evaluation protocol of prior work, CREDO raises Overall accuracy from 78.0% to 82.4% and Multiple accuracy from 55.0% to 65.6% on MultiTQ relative to the strongest published baseline in our comparison. On TimelineKGQA, it raises cross-KG accuracy from 30.9% to 36.4% relative to our reproduced baseline via direct transfer to ICEWS Actor under matched training and evaluation settings.
We study nonparametric regression using an over-parameterized two-layer ReLU neural network with fixed covariance-based channel attention, trained by gradient descent with both layers updated simultaneously. The spherical covariates follow an arbitrary fixed non-atomic distribution, and the target function lies in an interpolation space with spectral bias widely studied in the statistical learning literature. We prove that early-stopped training achieves regression risk $\cO(\eps_n^2)$, where $\eps_n$ is the critical population rate of the attention kernel $\Kattn$. Its positive eigenvalues are $λ_j^{s'}$, where $λ_j$ are those of the ReLU feature covariance kernel $K$ and $s'=2s+3$ for $s\ge1$. Under polynomial eigenvalue decay $λ_j\asymp j^{-2α}$ with $α>1/2$, the regression risk bound $\cO(\eps_n^2)$ specializes to the minimax optimal rate $\cO(n^{-2αs'/(2αs'+1)})$. Our analysis is based on two new technical results on the optimization and generalization of the attention network. First, the neural network function after every gradient descent iteration admits a decomposition into a function with bounded Reproducing Kernel Hilbert Space (RKHS) norm and a uniformly small residual. Second, local Rademacher complexity converts this decomposition into a regression risk bound. Kernel gradient descent on the same ball attains the same rate, while our theorem gives explicit sufficient width and auxiliary-sample conditions for simultaneous network training. Our analysis is further extended to covariates supported on $\RR^d$ through a novel normalization method. Under a sub-Gaussian covariate distribution, the attention network achieves a nearly optimal regression risk $\cO((\log n)n^{-s'(d+3)/(s'(d+3)+d)})$ without a density assumption.
Ruoyu He, Jisen Li, Anthony Ji, Changyi Yang, Zetian Li, Runduo Ma, Zhen Zhang, Zhongzhu Zhou
原文摘要
A full-duplex model should answer easy questions immediately, but hard questions often need stronger reasoning, which may require waiting. The challenge is to choose between these paths at the moment the model begins to speak. Recent systems learn related mode decisions through special-token generation inside the dialogue model, coupling routing to turn-taking and temporally annotated full-duplex training. We study a post-training alternative. We freeze MiniCPM-o 4.5, using only its speech input and output, and fit a logistic failure readout on hidden states already available at its native listen-to-speak transition. The resulting gate requires no additional model generation and exposes thresholds that can be retuned without changing the duplex policy. A probe fitted on text-input states also reads the states of spoken questions, including unseen human recordings. On the internal benchmark, selective handoff improves answer accuracy by 27.1 points. Compared with always using the expert, it reduces mean server time to first answer audio by 31.5 percent. It also outperforms matched random routing across external speech question-answering pools, as well as a zero-shot GPT-5.4 mini router that rates the transcribed question. The training recipe also transfers to a second duplex model family. These results show that an existing duplex model can be equipped to answer easy turns immediately and wait for stronger reasoning or fresher knowledge on the turns that need it, without further duplex training.
A companion study found that a complexity outlier rule shows where complexity concentrates but cannot tell a god unit from a complex algorithm. This paper pre-registered a test of one way to tell them apart: weight a unit's complexity by its coupling divergence, the data it hands on minus the data it receives, so that a leaf is set aside. Coupling was measured by resolving every call in 520 MLCQ repositories, 500 fresh Java repositories and 73 Qualitas Corpus systems. The claim fails: the primary, the pooled MCC of the weighted form minus complexity alone against MLCQ's human Blob, is -0.041 [-0.130, +0.054], no difference resolved and underpowered, and the resolved readings go against it. What the data settle is at method level. What MLCQ's reviewers call a Long Method is an absolute length: a constant of 26 lines or more, chosen on the explore half, reaches MCC 0.777 on the confirm half, +0.080 [+0.016, +0.145] over the best per-repository form; every reviewer with ten or more positives, on their non-leaf reviews, has a best cutoff between 17 and 34 lines; and nothing measured inside a method, its calls, variables, data flow or the cost of cutting it, refines the constant. The constant is a limit of human reading. The next form of the construct is the cost of bringing a long method under it, which is where the construct meets architecture.
Handwritten Urdu is hard to recognize because the cursive Nastaliq script has no fixed baseline and no clear character boundaries. We benchmark six systems on the Urdu Nastaliq Handwritten Dataset (UNHD): Tesseract (default configuration), a script-specific recognizer trained on printed Urdu (UTRNet-Large), two open-source vision-language models (VLMs), and two commercial VLMs. All six systems are evaluated without fine-tuning on 2,465 single-line images (Part A), and the four VLMs receive the same fixed prompt (decoding uses temperature 0 where configurable, except for Claude Sonnet 5, whose decoding cannot be configured). Both commercial VLMs reach lower error than the printed-Urdu recognizer and the open-source VLMs. We do not compare against handwriting-trained recognizers, which would likely do better in-domain; we characterize zero-shot behavior, not state of the art. We also describe three distinct failure modes among the ranked systems. In one of them, a commercial model explicitly declines to transcribe a few very hard-to-read images instead of guessing. In a second study (Part B), VLMs correct the output of a separate OCR system using both the source image and the noisy text. Once separated from each model's own reading ability, the benefit is small on Tesseract's output (0.7% to 7.4%) and larger (17.9% and 21.9%) only for the two commercial correctors given UTRNet's text. It appears to depend on how legible the supplied text already is, and a weaker corrector can make a more legible text worse.
Bayesian causal models explain how expectations change, but they leave a further question open: how does the learner acquire the causal content whose probability changes? I argue that updating alone leaves three conditions of causal judgment unexplained: experiential grounding, general organization, and belief. Hume links these conditions through a felt customary transition, which arises from discriminating resembling cases and supplies the idea of necessary connexion. His arguments against deriving necessity directly from a single interaction or from an act of will support this account of the source. Contemporary causal-theory learning uses perceptual and agency cues to acquire an intervention role. The resulting disagreement is about what that role explains about experienced causal necessity, not about whether the model learns anything. Three elementary constructions distinguish acquisition from assisted execution, lost distinctions from failed access, and posterior agreement from usable deployment. Sampling theories and stable-belief accounts supply further components beyond updating. For models that learn about external events solely through a text corpus and written reports, the evidence is report-mediated rather than firsthand; whether their own transitions are felt is a separate, open question, and the update rule settles neither matter. The paper identifies what a complete explanation must connect: the acquisition of causal content, its availability for general use, and its role in belief. Correct posterior prediction alone establishes none of these connections.
Depression affects millions of Bengali speakers across South Asia, yet automated detection systems have relied almost entirely on datasets translated from English. Such datasets cannot capture the emoticons, romanized Bengali words, and elongated spellings that characterize real expressions of distress. We introduce BanglishDepYT, a Bengali depression dataset collected natively from YouTube, comprising 191,000 unlabeled comments and 1,000 samples verified through dual bilingual annotation. We propose EPH (Emotion-Preserving Hybrid preprocessing), which converts emoticons to semantic tags, preserves Banglish tokens, protects negation words, and normalizes emotional elongation. A key finding is that preprocessing choice substantially affects performance on informal text: standard cleaning reduces the Banglish Survival Rate to zero (SVM F1: 0.7062), while EPH preserves signals (CESS: 0.9996 vs. 0.6694) and raises SVM F1 to 0.8570, an improvement of 0.1508. We further apply domain-adaptive pretraining (DAPT), continuing SahajBERT pretraining on unlabeled in-domain comments. DAPT-SahajBERT achieves the highest mean F1 (0.8161) and MCC (0.6272), significantly outperforming the baseline (F1: 0.7805, p = 0.0013, d = 1.450) and matching MuRIL (p = 0.589). These results suggest that targeted domain pretraining is a competitive alternative to large-scale multilingual pretraining. LLM-assisted annotation using Qwen2.5-7B achieves substantial human agreement (kappa = 0.611), enabling pseudo-labeling of 96,071 comments. Semi-supervised training improves AUC from 0.8499 to 0.8731 (p = 0.044), offering a scalable path to expanding annotation. All code, models, and datasets are publicly released to support reproducible research.
Purpose: We compared two multimodal System 1 decision models, Clef 27B and d1 3B, with a traditional LLM (Qwen 3.8-27B) on seven radiology tasks covering text, radiographs, CT and MRI. Methods: Three models were compared: Clef, Qwen3.8-27B and d1. Three tasks were interpretation of text and 4 were detection of finding in images. The tasks were: 1. finding presence of finding in CT-RATE chest CT reports; 2. Presence of finding in PadChest chest radiograph reports; 3. the same 18 findings on CT-RATE chest CT but 3D volumes; 4. MR safety status of devices from AccessGUDID device records; 5. line and tube position on RANZCR CLiP chest radiographs; 6. motion-artifact severity on MR-ART T1-weighted brain MRI and 7. sequence identification on BraTS 2021 brain MRI. The primary endpoint was AUROC. There was no fine-tuning of the models on the tasks. Results: All models completed all tests without a failed request. All retrieved findings from reports almost perfectly (AUROC >0.95). Clef ranked higher than its Qwen backbone on CT-RATE volumes and RANZCR lower on MR-ART. The two did not differ on BraTS, GUDID or PadChest, and differed by 0.002 on CT-RATE reports. On BraTS, Clef correctly identified 90 of 100 FLAIR volumes while Qwen never chose it. d1 ranked lower than Clef on six of seven tasks. No model performed well on motion severity. Clef was faster per request than Qwen on tasks with many questions per input. d1 used 3 to 9 times less energy per request than Clef. Conclusions: System 1 models can perform important tasks in radiology with high accuracy, at lower compute cost than LLMs used in the usual way. When an LLM is constrained to a single answer token, its cost and performance is similar to System 1 models unless multiple questions are asked. A 3B System 1 model was close to the 27B models on report text but not on images. No model performed well on motion severity.
We give a deterministic $O(m)$-time algorithm for exact modular subset sum over every modulus $m$. The input gives distinct residues and their multiplicities. The algorithm reports all reachable residues and answers one target query. It returns a witness when the target is reachable. The algorithm uses $O(m)$ auxiliary words on an arithmetic word-RAM. This improves Potępa's deterministic $O(m\log m\,α(m))$-time bound under the same input convention. The running time matches the cost of returning a length-$m$ Boolean array. We represent reachable residues as runs along cycles of repeated addition. Newly reached residues pay for scans of partial runs. Packed cycle flags make rebuilding take linear total time at the original modulus. A theorem on subset sums of distinct units limits the number of adaptive boundary batches to $O(m^{3/4})$. Radix sorting their $O(m)$ total keys also takes $O(m)$ time.
Deep Research (DR) systems extend large language models into autonomous agents that decompose complex information needs, iteratively search the web, and synthesize evidence to answer challenging questions. However, their multi-stage execution may introduce robustness risks, as variations in user questions or intermediate search inputs may propagate across research stages and lead to incorrect answers. Existing evaluations focus on task-solving capabilities or robustness against external evidence disturbances, leaving intrinsic weaknesses in DR execution insufficiently explored. Thus, we conduct an empirical study of 1,932 execution trajectories from three representative DR systems and identify nine recurring defect categories across subquestion formulation, online search, and final reasoning. Guided by these findings, we propose EMBERS, an empirically motivated, defect-guided fuzzing framework for robustness testing of DR systems. Starting from correctly answered seed questions, EMBERS generates stage-targeted test inputs using defect-guided mutation, web-search-assisted fidelity assurance, and diversity-aware seed maintenance. Experiments across 16 framework--model--dataset settings show that EMBERS achieves 3.21 times the failure-discovery throughput of the strongest baseline under identical testing budgets. Human and LLM-assisted evaluations demonstrate that most generated inputs preserve the original information need and necessary contextual constraints. Moreover, model fine-tuning experiments show consistent-answer performance improvement, highlighting their value for robustness testing and model improvement.
In an era defined by the explosive growth of data and rapid technological advancements, Multimodal Large Language Models (MLLMs) stand at the forefront of artificial intelligence (AI) systems. Designed to seamlessly integrate diverse data types-including text, images, videos, audio, and physiological sequences-MLLMs address the complexities of real-world applications far beyond the capabilities of single-modality systems. In this paper, we systematically sort out the applications of MLLM in multimodal tasks such as natural language, vision, and audio. We also provide a comparative analysis of the focus of different MLLMs in the tasks, and provide insights into the shortcomings of current MLLMs, and suggest potential directions for future research. Through these discussions, this paper hopes to provide valuable insights for the further development and application of MLLM.
Chaitanya Patel, Yusuke Kato, Rin Takahira, Kazuki Kozuka, Juan Carlos Niebles, Ehsan Adeli
原文摘要
Understanding holistic 3D human motion and the surrounding environment from egocentric video is fundamental for applications in AR/VR and robotics. Although these tasks are inherently coupled, existing methods treat them separately. In particular, egocentric motion reconstruction methods typically assume access to metric, gravity-aligned camera trajectories from specialized inertial sensors. Recovering such trajectories from monocular egocentric video remains a significant challenge due to rapid head movements and motion blur. We present EgoMo3R, a joint optimization framework that simultaneously reconstructs 3D human motion and scene geometry from casual egocentric video. It leverages the robustness of low-level visuo-geometric primitives, including monocular depth, optical flow, and point tracks, integrated with a diffusion-based human motion prior via first-order optimization. We propose an alternating optimization scheme that establishes a synergistic feedback loop: scene reconstruction provides physical grounding and trajectory conditioning for the motion prior, while the motion prior regularizes the camera trajectory and anchors the scene in a metric, gravity-aligned world frame. We evaluate EgoMo3R on the highly dynamic sequences of the EgoExo4D and Nymeria datasets and show that it outperforms specialized baselines in both 3D scene reconstruction and egocentric motion estimation.
Computational aesthetics is a field at the intersection of computer science, artificial intelligence, aesthetics, and other disciplines. As AI has developed, the field has formed relations of inclusion, intersection, or support with computational art, AI art, AI aesthetics, AI-generated content (AIGC), computational creativity, affective computing, and neuroaesthetics. Without distinguishing scope, objectives, and methods, it may be narrowed to image aesthetic quality assessment or expanded into an umbrella term for AI's involvement in art. We analyze its internal structure and its relations with neighboring concepts, and propose a two-level scope: special computational aesthetics (SCA) and general computational aesthetics (GCA). SCA centers on aesthetic measurement and aesthetic process modeling. The former primarily corresponds to "Aesthetic Perception": computationally identifying and representing aesthetic attributes and predicting human aesthetic evaluations; the latter primarily to "Aesthetic Understanding": computationally describing, testing, and explaining how human aesthetic perception and judgment are formed. GCA extends this core to include "Aesthetic Generation," namely artistic generation guided by computational aesthetic models. Some computational aesthetic tasks can also use non-AI methods. Research in this scope using AI algorithms and models we define as AI aesthetics, which may involve the special core and the general extension. "Aesthetic Generation" intersects with AIGC, AI art, and computational creativity, but belongs to the general extension defined here only when computational aesthetic models substantively constrain, guide, or optimize artwork generation. We also draw on Feng Peng's discussion of the computability of beauty, aesthetic experience, and art as a theoretical reference for clarifying boundaries between computational aesthetics and related concepts.
Leibnizian ideas in foundations of physics has enjoyed a renewed interest. Cyclic string ontology in Wolfram Model multiway rewriting systems has been found to encode finite dimensional quantum systems where what are called as Leibnizian strings have been employed. It has then become important to classify strings in terms of their ability to correspond to a monadic interpretation à la Leibniz. The definition of Leibnizian strings are rather quite non-linear and for that purpose we classify non-Leibnizian strings for which there is a simpler criterion on maximal neighborhoods of characters. If a string is not non-Leibnizian then it must be Leibnizian. We found out three canonical forms of non-Leibnizian strings and noted that two of them are context-free languages whereas one is not.
Anticipating financial outcomes requires synthesizing evolving evidence and reasoning under uncertainty, making deep research systems promising tools for forecasting. Yet evaluating forecasting ability is challenging because historical backtests may reward memorized outcomes or information retrieved after the forecast date. Live forecasting avoids direct access to realized outcomes, but feedback arrives only after events resolve. To obtain timely feedback and assess historical scores, we introduce FinForesight, a diagnostic backtesting benchmark for financial deep research systems. Across the Chinese, U.S., and Hong Kong markets, the benchmark evaluates four forecasting tasks: valuation, return ranking, event direction, and future financial fundamentals. Two capability diagnostic tasks assess counterfactual reasoning and financial calculation, scored separately from forecasting performance. Our harness-compatible framework restricts retrieved evidence to information available at the forecast date. We compare pretraining-era, resolved post-cutoff, and live evaluations to test whether controlled backtests approximate future performance. Pretraining-era scores exceed post-cutoff scores in most temporal comparisons, and stronger retrieval controls narrow the gap between post-cutoff backtests and live results. Forecasting attribution and robustness analyses reveal how scores respond to retrieval policy and how predictions respond to prompt rewrites and numerical perturbations. The code and benchmark data are publicly available.
Sichao Li, Sai Ma, Xiyang Hu, Chudi Zhong, Tongliang Liu
原文摘要
A common assumption in chain-of-thought (CoT) oversight is that a visible reasoning trace makes model behavior easier to monitor. We argue that this assumption can conflate distinct properties along a chain linking a model's internal computation, reasoning artifacts, outputs, and monitor judgments. We identify four properties along this chain: reasoning visibility (whether a trace is accessible), mechanistic faithfulness (whether a trace reflects internal computation), mechanistic localizability (whether a compact causal mechanism can be isolated), and behavioral monitorability (whether a monitor can detect the target behavior). Two complementary case studies across three open-weight reasoning models show why these distinctions matter. In the security vulnerability reasoning study, probes and activation ablations provide evidence of internal representation and causal influence, but a monitor with access to the reference answer does not outperform the majority-class baseline in a held-out test. In authority-hint factual reasoning study, a monitor correctly distinguishes changed from unchanged answers, but seven mechanistic analyses do not isolate a compact causal mechanism. These results show that evidence for one claim does not establish the others. We therefore propose an evidential-chain reporting standard. Each claim carries a licensing statement that specifies its target property, access tier, and artifact, and labels each required link as supported, unsupported, or assumed. These assessments determine whether the conclusion is licensed, conditionally licensed, or not licensed.
Hong Nie, Zhangchi Gao, Yuqian Liang, Jun Yu, Bo Wang, Jianqing Sun, Shengping Liu
原文摘要
Speculative decoding accelerates vision-language model (VLM) inference without altering the output distribution, yet two factors limit its gains. Existing drafters either lack the image evidence required for visually grounded tokens or pay to condition on every visual token, although a short draft block typically depends on only a few image regions. Moreover, the value of verifying additional candidates varies with GPU load, so no fixed verification structure is efficient across serving conditions. We present Uni-LaSpec, a block-parallel speculative decoding framework that addresses both factors. To improve draft quality, Uni-LaSpec reuses the verifier's own attention: the query retained from the preceding verification ranks the cached, int8-quantized visual keys, and a cross-layer vote selects a few visual tokens that the drafter fuses through gated cross-attention. Acceptance-aware training further weights each draft position by its contribution to the accepted prefix. To allocate verification, a marginal-cost criterion selects the structure per request and per round: draft trees exploit idle verifier capacity under light load, whereas a dynamic head truncates linear drafts under heavy load. Across six VLMs from three model families and 17 text and multimodal benchmarks, selecting eight visual tokens retains at least 98.6% of the acceptance obtained from the full image at sub-millisecond selection cost. With greedy decoding at concurrency one, Uni-LaSpec achieves up to 5.11x text and 3.24x multimodal throughput over autoregressive decoding. Across six load levels and three GPU types, it outperforms autoregressive decoding in every setting and matches or exceeds the best fixed verification strategy in 123 of 144 settings; under bursty arrivals, it reduces mean end-to-end latency by 47.6% relative to linear verification.
Khandokar Md. Rahat Hossain, Souvik Das, Tahsina Hashem, Mohammed Eunus Ali, Tanzima Hashem
原文摘要
Accurate house price prediction is essential for real estate buyers, sellers, agents, and financial institutions. Traditional methods rely primarily on structured data, e.g., number of bedrooms, bathrooms, parkings, but they overlook rich contextual information embedded in textual descriptions and property photos. Previous deep learning approaches for house price prediction have fused multimodal data but they fall short to extract structured spatio-temporal and contextual features from texts and semantic visual information from images. In this paper, we introduce a novel multimodal framework that addresses these limitations by systematically integrating structured attributes, textual descriptions, and visual images of properties to achieve higher prediction accuracy while maintaining interpretability. We use a large language model to derive domain-specific spatio-temporal and contextual features from text, and an unsupervised vision-language approach to automatically group photos into meaningful classes like indoor, outdoor, mixed, or floorplan, that are commonly present in real-world house listings. These categories are created without manual labeling by aligning image clusters with carefully designed pseudo text prompts. We fuse all features into a unified representation and use an optimized regressor to predict prices. Experiments on a large real-world property dataset demonstrate that our method outperforms traditional methods and existing multimodal baselines. Ablation studies confirm that the novel textual feature extraction and categorized visual embeddings substantially contribute to more accurate and reliable price predictions.
Fieldable low-light perception systems require both semantic and geometric scene understanding under severe domain shift and embedded compute constraints. We present FAMDA, a foundation-model-assisted self-training framework for unsupervised domain adaptation (UDA) in multi-task dense prediction. The framework integrates segmentation and depth foundation models into a self-training paradigm to generate high-quality pseudo-labels for the target domain, distilling their generalization capabilities into a single lightweight student network. To connect the method to deployment, we instantiate the adapted student on a compact handheld low-light prototype with DomiNite imaging and Jetson Orin Nano embedded compute. We evaluate on standard synthetic-to-real multi-task UDA benchmarks for comparison with prior work and on our field-collected low-light dataset to validate daytime-to-low-light adaptation in the target application setting. Across these evaluations, the proposed approach achieves strong segmentation and depth accuracy with substantially lower memory use and inference latency than direct foundation-model inference, demonstrating a practical path toward embedded multi-task perception for low-light field systems.
Every weather forecast is uncertain, and decisions from flood warnings to energy trading depend on how well that uncertainty is quantified. Machine-learning (ML) ensembles now quantify it better than the best physics-based ensemble, yet two things about them remain opaque: whether their skill requires hundreds to thousands of accelerator-days of training, and how a network with a simple source of randomness produces a spread that follows the flow. Here we introduce Sonny-ENS, an ensemble of three 68.9-million-parameter networks with Monte Carlo dropout, trained on the continuous ranked probability score (CRPS) in 2.7 A40 GPU-days. On 2022, it has a lower CRPS than the European Centre for Medium-Range Weather Forecasts ensemble (ENS) on 96% of 570 targets and an 8.3% lower mean CRPS at lead times of 1-10 days. It outperforms the operational GenCast on 69% of the targets, and a 51-member, 15-day forecast takes 69 s on one GPU. Ablations identify the ingredients of this skill, and a Kolmogorov-Arnold sparse autoencoder shows that dropout injects uncertainty mainly in the tropics, whereas the spread grows where the atmosphere is baroclinically unstable (spatial correlation 0.91 with the Eady growth rate). Skillful probabilistic weather forecasts are therefore within reach of a single GPU server.
Changcheng Song, Yi Yang, Xiangyu Chang, Xiaoling Lu, Yufeng Liu
原文摘要
Fairness-aware classification seeks to balance predictive accuracy and fairness across demographic groups. Although surrogate losses facilitate the optimization of fairness-aware objectives, it remains unclear when minimizing these losses yields optimal decisions under the original fairness-aware risk. This paper investigates the statistical validity of surrogate-based fairness-aware classification under demographic parity and equal opportunity, considering both attribute-blind and attribute-aware settings. We introduce fairness-aware Fisher consistency and characterize it through conditions on the derivatives of the surrogate losses. The characterization shows that consistency depends not only on the individual loss functions but also on their interaction within the fairness-regularized objective. We further establish a calibration inequality that relates the excess fairness-aware risk to the excess surrogate risk, providing a quantitative connection between surrogate optimization and the original classification objective. Numerical experiments illustrate how deviations from the consistency conditions affect fairness-aware excess risk and how the regularization parameter influences the well-posedness of surrogate optimization. An application to consumer credit default prediction demonstrates that appropriately calibrated fairness regularization can substantially reduce demographic disparities while maintaining predictive performance.
The arboricity of a graph is the minimum number of forests required to cover all its edges. In this paper, we examine arboricity from a game-theoretic perspective and investigate cost-sharing in the minimum forest cover problem. We introduce the arboricity game as a cooperative cost game defined on a graph. The players are edges, and the cost of each coalition is the arboricity of the subgraph induced by the coalition. We study properties of the core and propose an efficient algorithm for computing the nucleolus when the core is not empty. In order to compute the nucleolus in the core, we introduce the prime partition which is built on the lattice of maximum-density edge sets. The prime partition decomposes the edge set of a graph into a non-prime set and a partially ordered collection of prime sets defined from minimal densest minors and their invariant precedence relation. Moreover, edges from the same set of the partition always have the same value in a core allocation. Consequently, when the core is not empty, the prime partition significantly reduces the number of variables and constraints required in the linear programs of Maschler's scheme and allows us to compute the nucleolus in polynomial time. Besides, the prime partition provides a graph decomposition analogous to the celebrated core decomposition and the density-friendly decomposition, which may be of independent interest.
Verifying that a packaged food is Halal requires reading ingredient declarations and E-number additives before goods enter certified warehouses. Dual-origin additives such as E471, E422, E441 and E920 are Mushbooh (doubtful) unless a botanical, microbial or certified-Halal qualifier appears in the same ingredient clause, so a plant-based qualifier on one clause must not clear an unqualified additive in another. We call a failure to respect this scope cross-clause modifier confounding and build a benchmark of 3,199 items (277 Open Food Facts records and 2,922 template-generated clause-swap items, labelled by an explicit 50-code rule base rather than by a certification authority) with a 540-item counterfactual suite. We propose Halal-NeSyNet, which couples a DistilBERT encoder with a 16-dimensional rule-derived symbolic vector through cross-attention and gated fusion, trained with an asymmetric false-Halal risk loss. Over five seeds, text-only TF-IDF baselines reach 86.25-86.71 % accuracy with 13.42-21.73 % false-Halal leakage and 6.33-26.00 % counterfactual triad consistency, whereas Halal-NeSyNet reaches 97.04+/-0.34 % and 99.67+/-0.50 %. Gradient boosting given the same symbolic vector reaches 96.25+/-1.24 % (+0.59 points from Halal-NeSyNet, 95 % CI [-1.11, +2.28]), so most of the gain comes from clause-scoped symbolic features and not from the network, and the risk loss did not help. Holding out whole template families, text-only models fall to 61.91-78.35 %, Halal-NeSyNet to 90.87+/-11.76 % and logistic regression with the symbolic vector to 91.11+/-2.88 %, with far more spread for Halal-NeSyNet. A three-lane receiving simulation auto-clears 49.33 % of cargo at 0.76 % green-lane contamination. Because labels and features come from one rule engine and 91.3 % of the items are templated, the results measure fidelity to a rule policy, not religious certification.
Spiking Neural Networks (SNNs) provide an event-driven paradigm for energy-efficient edge computing. However, operation-level sparsity in SNNs does not necessarily eliminate the hardware costs associated with redundant output neurons, including neuron-state maintenance, partial-sum (psum) accesses, and on-chip storage requirements. This paper presents SparseForge, a hardware-aware sparse SNN acceleration framework that jointly optimizes neuron pruning, sparse dataflow, and accelerator microarchitecture. At the algorithm level, we introduce a gated leaky integrate-and-fire (G-LIF) neuron to learn differentiable neuron masks with binary spike semantics. A dual regularization scheme is further developed to align pruning with both synaptic-operation reduction and psum buffer capacity reduction. At the dataflow level, we design a sparse Column-Wise Product (CWP) dataflow with a variant of incremental spike encoding to improve weight-access regularity and reduce sparse-index overhead under dynamic spike activity. At the architecture level, we develop a mask-aware sparse processing unit (SPU) that supports compressed spike decoding, sparse implicit im2col conversion, neuron-mask-gated psum updates, and decoupled sparse execution. Experimental results across multiple SNN backbones and datasets show that SparseForge retains 1.75%--77.99% of neurons and 0.16%--62.36% of synaptic operations, with a configurable accuracy--efficiency trade-off. The buffer-capacity-aware regularization reduces the required psum buffer capacity by up to 96.18\%. FPGA implementations on a Xilinx ZCU9EG achieve 296.94--619.49 GOPS/W across the evaluated workloads. Across matched network and dataset settings, SparseForge provides 1.24--4.17 higher FPS and 1.50--7.18 higher GOPS/W than representative FPGA-based SNN accelerators, while reducing EDP by 1.70--19.13, with benchmark-dependent accuracy--efficiency trade-offs.