Artificial Intelligence
DI-Bench: Systematically Generating In-Domain Data Intelligence Benchmarks for Enterprise Agents
Jiangyun Zhang, Kristen Surrao, Torpong Nitayanont et al. (+10 more)
Evaluating enterprise
agents on domain-specific
benchmarks is critical, yet public
benchmarks rarely evaluate whether
agents can integrate business knowledge with analytical computation, and constructing such
benchmarks manually is costly. We present DI-Bench, a pipeline for generating realistic
benchmarks for data intelligence (DI), the practice of extracting insights from large volumes of enterprise data. To emulate realistic DI tasks that require both computation and knowledge
retrieval, DI-Bench builds an artifact linkage graph over data tables, dimensions, metrics, and documents to form questions involving structured data and associated knowledge. Ground truth answers are derived via query execution, followed by LLM question generation and validation. Applied to two public
datasets, the pipeline produces a 731-task
benchmark covering knowledge
retrieval, analytical computation, and rule-grounded
reasoning. To show the discriminatory capability and difficulty of the
benchmark, we evaluate four models, revealing a substantial finding: models achieve only 32% accuracy when doing computational tasks where retrieved business rules modify the computation.
Read the full abstract ▾
All authors Jiangyun Zhang, Kristen Surrao, Torpong Nitayanont, Yupei Zhang, Roopali Singh, Zhiyu Chen, Julia Huang, Zhou Tang, Shayan Ali Akbar, Omar Alonso, Erwin Cornejo, Yuan Li, Yi Zhang
Artificial Intelligence
Inference-Time Graph Engineering for Multi-Agent LLM Workflows
Katherine Tieu, Dongqi Fu, Yinglong Xia et al. (+3 more)
Recent multi-
agent LLM systems increasingly rely on graph-structured communication to coordinate specialized
agents. We revisit multi-
agent orchestration from a graph-engineering perspective: rather than optimizing a static
topology, we synthesize a task-conditioned temporal workflow graph that jointly specifies
agent connectivity and edge-level communication semantics. We introduce ReActNet, a training-free framework that compiles a query and a set of role-specialized
agents into a sequence of directed communication graphs. Each graph snapshot corresponds to one
reasoning stage, and each edge carries a natural-
language instruction specifying the message that a source
agent should provide to a target
agent. The compiled temporal graph is then executed through structured message passing:
agents update their
reasoning states by integrating their previous states with messages from controller-assigned neighbors, and a final aggregator synthesizes the resulting states into the answer. This design separates graph compilation from graph execution, making multi-
agent coordination explicit, inspectable, and task-conditioned without requiring reinforcement learning or gradient-based
topology optimization. Across knowledge
reasoning, mathematical problem solving, code generation, and
GAIA-style assistant tasks, ReActNet consistently improves over fixed-
topology and learned-
topology baselines while maintaining competitive inference cost. These results suggest that effective multi-
agent orchestration depends not only on which
agents communicate, but also on engineering executable workflow graphs that encode when, why, and how information should flow during
reasoning.
Read the full abstract ▾
All authors Katherine Tieu, Dongqi Fu, Yinglong Xia, Hong Li, Hong Yan, Jingrui He
Information Theory
Moments of crosscorrelation demerit factors of binary sequences
Daniel J. Katz and Harmony M. Vargas
Families of sequences with low mutual aperiodic crosscorrelation assist the design of systems for multi-user asynchronous communications and multiple-input multiple-output radar. The crosscorrelation demerit factor of a pair of sequences is the sum of the squared magnitudes of their crosscorrelation values at every shift when the sequences are normalized to unit Euclidean norm, and the merit factor is the reciprocal of the demerit factor. For each positive integer $\ell$, we endow the $2^{2 \ell}$ pairs of binary sequences of length $\ell$ with uniform probability measure and study the distribution of their crosscorrelation demerit factors. Sarwate showed that the mean value is always $1$ regardless of length $\ell$. We develop a method for finding an exact formula for the $p$th central moment (for any positive integer $p$) as a function of $\ell$. Formulae for the variance and third central moment ($p=2$ and $3$) are then obtained by hand calculations, while the fourth through sixth central moments are obtained by computer-assisted calculations. Our theory also shows that all the central moments must be strictly positive for $p\geq 2$ and $\ell \geq 3$.
Read the full abstract ▾
Machine Learning
RAPTOR: Role-Aware Private Training for Mixture-of-Experts
Duc Dm, Khai Le-Duc, Nguyen Do et al. (+18 more)
Differentially private (DP) fine-tuning methods treat sparse Mixture-of-Experts (MoE) models as a single dense block, ignoring that shared layers see all data while experts only see routed records. We identify and formally characterize three resulting failure modes: global clipping suppresses expert gradients, batch-level normalization dilutes sparse expert updates, and fixed
privacy noise degrades signal-to-noise ratio on low-load experts. We introduce RAPTOR - a Role-Aware Private Training framework, which alternates shared and expert optimization and targets each failure directly, using expert-specific clipping and noise together with a public expected-owner denominator and a count-independent update schedule that avoids conditioning on private, realized expert counts. We prove the resulting mechanism satisfies $(\varepsilon,\delta)$-DP: because each record is assigned to exactly one owner expert, per-expert mechanisms within a layer compose in parallel, so updating all $E$ experts costs no more, in
privacy terms, than updating one, with shared and expert streams composing sequentially across training. We further derive a bias-variance decomposition of the public-denominator
estimator showing its bias grows predictably with routing imbalance, yielding a
privacy-free rule for selecting which layer to protect from routing
entropy measured on a small public corpus. Experiments on Switch
Transformer and OLMoE fine-tuning across GLUE tasks, and on DeepSeek-VL2-Tiny, show consistent gains over standard DP baselines across several
privacy levels ($\varepsilon$), with the largest margins typically at the tightest budgets. Code and models are publicly available: https://github.com/leduckhai/RAPTOR
Read the full abstract ▾
All authors Duc Dm, Khai Le-Duc, Nguyen Do, Minh Son Hoang, Florent Draye, Thai Hoang, Hoang Phuong Dam, Jiarui Liu, Chris Ngo, Terry Jingchen Zhang, Anh Le Duc Tran, Nhat Do Minh, Minh Ngoc Le, My T. Thai, Ran Xu, Silvio Savarese, Mona Diab, Bernhard Schölkopf, Zhijing Jin, Huy L. Nguyen, Daeyoung Kim
Human-Computer Interaction
Will My Assistant Remember My Allergy? What Personal LLM Assistants Forget When Conversation Memory Is Compressed
Lichen Zhu, Yueqian Lin, Yiheng Wang et al. (+3 more)
Personal LLM assistants (health companions, elder-care
agents, accessibility aides) are judged by what they remember about a person: a medication or an allergy mentioned in passing and needed days later.
Privacy pushes them on-device, where a month of conversation can outgrow the model's own weights, so an eviction policy must decide what the cache forgets.
Benchmarks report that eviction keeps such facts at a 20% budget, but they compress a
prompt that already contains the user's future question, foresight no cache-reusing assistant has. Hide the question until after compression and the advantage vanishes: on PA-Bench, 100 assistant conversations we construct, an allergy mentioned in passing survives to the question that needs it 0–1% of the time, against 97% with full
memory. The cause is the budget, not the scorer: none of the training-free policies we evaluate ranks the fact high enough, and the budget that would keep it is too large to bother compressing. A compressed cache is an inference-reuse mechanism, not a persistence layer: safety-critical facts need an auditable episodic store alongside it, and an interface that asks rather than invents.
Read the full abstract ▾
All authors Lichen Zhu, Yueqian Lin, Yiheng Wang, Yudong Liu, Hai "Helen" Li, Yiran Chen
Machine Learning
Data Scout: Targeted Web Crawling for Domain-Specific Pretraining Corpora
Chirag Garg, Eelaaf Zahid, Farhan Ahmed et al. (+3 more)
The dominant approach to building domain-specific pretraining corpora is to filter large web archives such as CommonCrawl. This works well for popular domains but breaks down for specialized ones, where relevant content is sparse and often beyond the reach of popularity-driven crawlers. We present Data Scout, which inverts this: instead of filtering an archive, it directs a targeted crawl. An LLM expands a root topic into a taxonomy and thousands of search queries; the returned URLs (seeds) are grouped by subdomain and screened with a user-supplied classifier (the probe), admitting each subdomain on the basis of a small sample. This works because relevance has a sharp boundary at the subdomain level: in mathematics, a page is 21x more likely to be relevant than one on a sibling subdomain. With the FineMath classifier as the probe, 21.9% of crawled pages are high-quality math content, 70x the 0.31% rate from filtering a comparable web sample, so the crawl wastes far less effort. But the payoff is not just efficiency: 63.2% of these pages are missing from CommonCrawl altogether, yet just as useful for training. Continued pretraining of Llama-3.2-3B on 1.9B Data Scout
tokens matches FineMath corpus on GSM8k. Because the probe is the only domain-specific component, Data Scout can in principle apply to any domain with such a classifier.
Read the full abstract ▾
All authors Chirag Garg, Eelaaf Zahid, Farhan Ahmed, Jay Pankaj Gala, Eric Butler, Heiko Ludwig
Game Theory
Utility-Driven Spatial Data Sampling for UAV-Assisted Scientific Smart Farming
Keiwan Soltani and Sajal K. Das
Large smart-farming deployments generate continuous scientific data from spatially distributed sensors, including soil, humidity, temperature, crop-health, and pest-related measurements. In vast agricultural fields, however, an energy-constrained unmanned aerial vehicle (
UAV) often cannot collect data from every sensor during each mission. Existing
UAV-assisted collection methods typically optimize coverage, route length, data volume, or freshness, but they do not always distinguish between data that is merely available and data that is scientifically valuable. This poster introduces a utility-driven spatial sampling framework for
UAV-assisted smart farming. The field is partitioned into grid
cells, each sized according to the
UAV ground coverage range. After an initial exploration phase, each
cell receives a scientific utility score based on freshness, redundancy, anomaly likelihood, and model uncertainty. The
UAV then selects and visits a subset of high-utility
cells under battery and return-to-base constraints. The proposed framework reframes
UAV-based collection as adaptive scientific data management rather than exhaustive sensing.
Read the full abstract ▾
Hardware Architecture
Interface-Aware KV Cache Quantization for Dense On-Chip NVM in Long-Context LLM Decoding
Jiahao Zheng, Yifan Qin, Xiaobo Sharon Hu et al. (+1 more)
The key-value (KV) cache is the dominant
memory bottleneck in long-context large
language model (LLM) decoding: every step reads it entirely, so decoding is
memory-bandwidth bound. Holding a quantized KV cache in dense on-chip non-volatile
memory (NVM) removes the off-chip transfer. Existing KV quantization methods, however, were designed for GPU-style
memory systems: KIVI attaches per-group metadata, adding about 25% to the stored KV cache; KVQuant keeps sparse full-precision outliers that a dense array cannot hold in place. This paper examines what these structures cost when the KV cache resides in NVM behind fixed-range converters, and designs a quantization scheme matched to that interface. The architecture stores the quantized KV cache in dense on-chip NVM, uses a small static analog crossbar only for the fixed rotation, and keeps
attention in on-chip digital logic. A randomized rotation and per-vector normalization give every coordinate the same range, so one fixed codebook for keys and one for values, each shared across all
tokens of the corresponding tensor type, serve the entire KV cache. The codebook thresholds are programmed once as the read converter's reference levels, enabling fixed-range digitization with no per-
token converter reconfiguration. Dequantization is a sixteen-entry lookup and one norm multiply; the only per-vector metadata is one scalar, about 3%. Across models from 3B to 14B and contexts to 32k
tokens, the four-bit KV cache maintains accuracy under storage and crossbar noise simulated at realistic device levels. KIVI and KVQuant remain more accurate in software; the advantage of our format lies at the
memory interface: 3.1-3.6x lower KV read energy than both mapped to the same NVM, and 8x lower metadata overhead than KIVI. The contribution is a KV quantization co-designed with the NVM
memory interface rather than a new accuracy record.
Read the full abstract ▾
All authors Jiahao Zheng, Yifan Qin, Xiaobo Sharon Hu, Yiyu Shi
Systems and Control
AgentHomeID - Agent-based modelling of building stock transformation: A multi-scale framework for policy assessment and infrastructure planning
Helen Ganal, Sarah Becker, Sascha Holzhauer et al. (+4 more)
Decarbonising the building sector is central to meeting
climate targets, yet existing models rarely capture the interaction between system-level transformation dynamics and heterogeneous individual investment decisions. This work presents AgentHomeID, an
agent-based model of building stock
evolution in which owner behaviour, techno-economic constraints, and regulatory frameworks are represented explicitly at the level of individual buildings and their owners. The model differentiates owner-occupiers, private landlords, and institutional owners, using willingness-to-pay (WTP) parameters estimated from empirical decision-maker studies, and operates on both representative building archetypes and real building data derived from geographic information systems (GIS). We demonstrate this versatility across three applications. At national scale, scenario analysis for Germany to 2045 shows that removing binding renewable heating requirements substantially raises final energy demand even where envelope refurbishment is unchanged, and that subsidy allocation and investment activity diverge sharply across owner types and income quartiles, with the lowest quartiles persistently underinvesting. At regional scale, bottom-up simulation for a German distribution grid planning region yields spatially concentrated heat pump uptake at NUTS-3 level that differs from aggregated top-down projections in both magnitude and spatial distribution. At urban block level, the same simulations resolve substation-level load heterogeneity and show that integrated system peaks driven by heat pumps, electric vehicles, and photovoltaics do not coincide with individual technology peaks. Across all three scales, owner heterogeneity and local structure materially shape transition pathways, indicating that they should be represented explicitly in models used for policy assessment and infrastructure planning.
Read the full abstract ▾
All authors Helen Ganal, Sarah Becker, Sascha Holzhauer, Thilo Glißmann, Friedrich Krebs, Martin Braun, Philipp Härtel
Computer Vision
AAMBERS-UAV: Acquisition-Aware Multimodal Backbone Evaluation and Ranking for UAV Weedy Rice Segmentation
Tarek Rahman, Nazim-E-Alam, Md Kishor Morol et al. (+1 more)
UAV image collections contain spatially and temporally related frames, yet semantic-
segmentation benchmarks commonly split them at image level. Such splitting can place samples from one acquisition in both model development and testing, obscuring transfer to a genuinely new survey. Using the 734-sample WeedyRice-RGBMS-DB, we fix a 124-image target-acquisition test set and compare two protocols with identical train, validation, and test counts: target-held-out, which excludes the target acquisition from development, and target-exposed, which admits its remaining images. SegFormer-B0 is evaluated with RGB, four-band multispectral (MS), and seven-channel RGB+MS input over two fixed-split seeds. RGB is strongest under complete acquisition holdout ($0.7317\pm0.0201$ IoU), whereas RGB+MS becomes strongest after target exposure ($0.7822\pm0.0269$). A fixed-split U-Net/ResNet18 replication confirms positive exposure gains for all three inputs, but retains RGB as the best modality under both protocols. Acquisition exposure therefore increases measured performance across both evaluated backbones, while its effect on modality ranking is architecture-dependent. A supplied-split audit reveals strong near-sequential dependence, and corruption tests show that early fusion is substantially more sensitive to RGB–MS displacement than to moderate radiometric scaling. These results support acquisition-aware same-test evaluation as a necessary complement to ordinary image-level splitting in
multimodal UAV benchmarks. The code and supporting the findings of this study will be publicly released upon acceptance of the paper.
Read the full abstract ▾
All authors Tarek Rahman, Nazim-E-Alam, Md Kishor Morol, Jannatun Noor
Computer Vision
GeoContext: One Context Ladder, Two Failure Modes in Vision-Language Geolocation: Flat Reliance on User-Provided Location Context and False Confirmation of Location Claims
Yifan Zhang and Kai Wang
Visual geolocation
benchmarks typically ask a model where an image was captured without accounting for the location context that users often provide. We introduce GeoContext, a resource supporting two complementary tasks: GeoHint, open-ended localization given a true but coarse location hint, and GeoVerify, binary verification of whether an image was taken within 150 m of a claimed place. GeoContext constructs a context ladder by stratifying nearby reference points according to distance and referenceability, allowing the image to remain fixed while the supplied context varies. The
benchmark covers 109 sites in 30 cities and evaluates five
vision-
language models using 21,933 GeoHint responses and 6,270 GeoVerify responses. Our evaluation reveals three main patterns. First, hint repetition varies by only 1.5 percentage points across referenceability tiers and by less than 3 points across distance bands, while the resulting localization error increases steadily with hint distance. Second, behavior depends strongly on no-context performance: at sites with low no-context accuracy, the median ratio between localization error and hint distance is approximately 1.00, whereas at higher-accuracy sites it ranges from 0.24 to 0.69. After correcting for bias introduced by the site grouping procedure, only one of the five models retains a negative accuracy estimate when given a nearby hint. Third, in GeoVerify, no model reaches d' = 1 for decoys immediately beyond the 150 m tolerance. Model rankings also change when sensitivity is separated from response bias, and 83.8% of false acceptances are reported with confidence of at least 0.8. We release the
benchmark, construction pipeline, audit decisions, and scoring code.
Read the full abstract ▾
Performance
RAGMark: A Comprehensive Framework for Benchmarking Retrieval-Augmented Generation Systems
Zlatan Feric, Amir Taherin, Bin Ren et al. (+3 more)
We present RAGMark, a modular benchmarking framework for advanced
Retrieval-Augmented Generation (RAG) systems targeting small-scale multi-GPU environments. RAGMark evaluates diverse RAG components, including retrievers, vector databases,
prompt-processing methods, and generator models, while collecting detailed per-stage metrics such as latency, GPU utilization,
memory consumption, power usage, time to first
token (TTFT), throughput, and answer quality. The framework is highly extensible, separating RAG stages, timing, and resource monitoring into modular components, and is designed to efficiently sweep large configuration spaces while minimizing repeated model and database initialization overhead. Using RAGMark, we characterize five RAG workloads on open-domain QA
datasets across varying
retrieval depths, model scales, reranking, compression methods, and vector database configurations. We show that while autoregressive generation dominates latency in naive pipelines, context-reduction techniques shift bottlenecks across compute,
memory bandwidth, and preprocessing stages. Reranking and compression produce compounding benefits: reranking reduces compression workload itself, while both jointly reduce prefill and KV-cache traversal costs, lowering energy consumption by up to 66%. We further observe strong cross-stage interactions, where small upstream context reductions cascade through downstream latency,
memory traffic, and energy consumption. The RAGMark source code is publicly available at: https://github.com/zferic/RAGMark.
Read the full abstract ▾
All authors Zlatan Feric, Amir Taherin, Bin Ren, Yanzhi Wang, Jennifer Dy, David Kaeli
Information Retrieval
What Price Fairness? Evaluating Energy - Fairness - Accuracy Trade-off in Recommender Systems
Abhirup Mitra, Oleg Lesota and Antonela Tommasel
Fairness-aware recommender systems aim to mitigate systematic imbalances in recommendation outcomes, including how visibility, relevance, and opportunities are distributed among users, items, and providers. However, these systems are usually evaluated in terms of accuracy and
fairness alone, while their computational and environmental costs remain largely invisible. This omission matters because
fairness interventions may affect the cost of recommendation in different ways. Training-time methods modify model optimization, post-processing methods add computation at inference time, and both may depend on the model,
dataset, hardware, and deployment setting. We examine whether provider-side
fairness in recommendation comes with a measurable green cost. We compare in-processing, graph-level reweighting and post-processing interventions across multiple models, two
datasets, and two hardware settings. We measure recommendation quality, provider-side exposure, and energy consumption separately across training and inference stages. Our results show that the green cost of
fairness is not uniform, post-processing shifts cost to repeated serving, while in-processing and graph-level methods avoid re-ranking overhead but vary substantially across models,
datasets, and hardware. Findings call for evaluating
fairness-aware recommendation as a three-way trade-off between accuracy,
fairness, and computational cost.
Read the full abstract ▾
Artificial Intelligence
From Monolithic Blending to Agentic Orchestration: Dynamic Response for Conversational Assistants at Scale
Cen Mia Zhao, Peng Wang, Chuan Shi et al. (+6 more)
Conversational assistants can blend
retrieval, action selection, escalation, and wording in a single model path, or separate those roles. We report a production migration of a customer-support assistant at a large accommodation marketplace (millions of conversations per month, 11
languages, 10-second P90). Dynamic Response (DR) replaces a single Qwen3-235B-A22B blended responder with a bounded ReAct orchestrator over typed tools plus a smaller generator that writes from a backend-validated context contract. Because the migration also changed
prompts,
alignment, and serving, we attribute each effect to its cause and claim as architecture effects only those measured on identical replayed turns: typed entity selection moves the reservation selector to a precision-first operating point (precision 8.3% to 89.1%, recall 75.2% to 67.3%), and typed action IDs with a membership check remove observed structured-action
hallucination (2.14% to 0.0%). A low-ramp A/B test reproduces the replay escalation reductions: hard-escalation responses fall from 5.60% to 3.08% and soft-escalation responses from 9.56% to 2.49%, while production handoff volume holds roughly steady; self-solve is directional (+5.1 points, 95% CI [-2, +12]). Serving optimizations cut orchestrator P90 latency from 3.87s to 2.24s on a GPU footprint reduced by roughly one-third, and self-hosting reduces estimated annual model-serving cost by more than an order of magnitude.
Read the full abstract ▾
All authors Cen Mia Zhao, Peng Wang, Chuan Shi, Yufeng Zhang, Ying Lyu, Wanmeng Ren, Robert Xue, Claire Na Cheng, Yashar Mehdad
Computation and Language
CrisisKD: Five-Stage Knowledge Distillation for Aspect-Level Sentiment and Emotion Analysis in Crisis Discourse
Marko Haralović, Onat Akca, Salih Eren Yücetürk et al. (+2 more)
Identifying the target of emotional words or phrases in crisis situations, especially health-related ones, is important for understanding public concerns across cultural and linguistic contexts. We propose CrisisKD, a five-stage teacher–student knowledge
distillation framework for aspect-level sentiment and
emotion analysis on unannotated social media data. A teacher LLM generates aspect-level labels and
reasoning traces that supervise a smaller student model across aspect extraction, syntactic parsing, opinion extraction, sentiment classification, and
emotion classification. Using this framework, we construct and release a
dataset containing 50,615 aspect-level labels, together with the annotation and fine-tuning scripts as open-source resources. The resulting student supports end-to-end ABSA and
emotion detection at substantially lower inference cost than the teacher. On a manually annotated 500-tweet gold set, the 5-task Qwen2.5-7B student improves over the untuned model by 7.9 F1 points on aspect extraction, 17.0 points on
emotion accuracy, and 6.5 points on sentiment accuracy. On the external ABEA
benchmark, CrisisKD improves the same-model Qwen2.5-7B ICL baseline by 2.8 F1 points on ATE and 3.8 F1 points on joint ATE+AEC.
Read the full abstract ▾
All authors Marko Haralović, Onat Akca, Salih Eren Yücetürk, Minsi Li, Mariët Theune
Databases
Concord: A Video Relational Algebra for Cross-Modal Query Optimization
Sultan Muratbek, Charisse Ivana Yeung, Chanwut Kittivorawong et al. (+1 more)
Semantic
video queries let users embed natural
language prompts and use
multimodal large
language models (MLLMs) to interpret the
video. Such queries are increasingly popular for querying
video data. However, their expressiveness comes at a steep cost: an MLLM may process hours of media to return only seconds of relevant output, making naive execution slow, expensive, and inaccurate. We propose Concord, a system for expressing and optimizing semantic
video queries. We makes three contributions. First, we introduce
Video Relational
Algebra (VRA), a nested
algebra over
videos, transcripts, frames, and object tracks that captures common semantic
video operations. Second, we derive a set of approximate optimizations that rewrite VRA queries to reduce MLLM usage while improving result quality. For narrated
video, Concord either processes transcripts instead of
video or uses them to identify
video clips for MLLM processing. For cross-camera queries without narration, detection and tracking replace a whole-
video MLLM join with a track-level relational join. Third, we evaluate Concord on real-world
videos. Across 4.59 hours of soccer broadcasts and 3.92 hours of lectures, transcript-to-
video queries send only 5.32% and 2.47% of source-
video duration to the MLLM and reduce MLLM cost by up to 87%. In two five-second highway clips with 18 manually adjudicated cross-camera vehicles, a Detect-Track-Join query improves F1 from .364 to .813 while making no MLLM calls. See our project at https://concord-db.github.io/.
Read the full abstract ▾
All authors Sultan Muratbek, Charisse Ivana Yeung, Chanwut Kittivorawong, Alvin Cheung
Sound
Emotion as a Distribution: Joint Valence-Arousal Probability Learning for Speaker-Independent Multimodal Emotion Recognition
Tingyi Lin, Wen-Ren Yang and Kuanwei Chen
Human
emotion is graded and frequently mixed, yet most
multimodal recognizers collapse it onto a single hard label. We argue the recognizer should instead expose a distribution over affective space. Our text+
speech system, alongside its categorical decision, emits a $9\times9$ probability
matrix over the Valence-Arousal plane, trained with a two-dimensional Gaussian soft target under a Kullback-Leibler/cross-
entropy objective, aimed at counseling support. Evaluation is strict: speaker-independent 5-fold leave-one-session-out IEMOCAP with rotating-session inner validation, headline metrics only on the held-out session. Within one fixed encoder-fusion-head pipeline we compare
Transformer and state-space (Mamba-1/2/3) backbones at matched depth and width, at two operating points ($T\approx550$, $T\approx2750$). The featured dual-head system reaches 73.0% $\pm$ 0.3 unweighted accuracy over three seeds (separate rerun: 72.1%), exceeding the
Transformer fusion baseline by 3.0 UA points (95% session-bootstrap CI [1.0,4.7]; significant under paired t-test and session-level bootstrap), with no latency or
memory advantage at these lengths; swapping the ~1M trainable front-end for frozen WavLM-Large features (learnable layer weights) lifts the same architecture to 76.6% $\pm$ 1.3. Pre-specified controls scope the claims honestly: simpler valence-arousal auxiliaries reproduce the classification lift within noise, and a dedicated regression head tracks the continuous ratings slightly better, so the head's specific value is the normalized affect distribution itself. That distribution recovers the circumplex: its center of mass tracks valence and arousal (CCC 0.66/0.66; predominantly between-class structure, weaker within-class tracking), and its
entropy is weakly but consistently linked to categorical rater ambiguity, not dimensional spread.
Read the full abstract ▾
Information Theory · Cross-listed from math.ST
A complete characterization of sequential testability and change detectability in i.i.d. models
Aaditya Ramdas
We give a necessary and sufficient condition for the existence of power-one sequential tests in an i.i.d. composite testing problem. A level-\(\alpha\) test with power one against every alternative exists if and only if the alternatives are separated from the null by a countable family of finite-block events. We provide other equivalent conditions using randomized fixed-sample tests, bounded finite-block scores, e-processes, reduced-filtration test supermartingales, and a countable cover whose finite-block weak-$*$ closed convex hulls are positively separated in total variation. As a bonus, the constructive proof yields tests have pointwise expected sample size \(O_Q(\log(1/\alpha))\). Exactly the same conditions also characterize i.i.d.\ change detectability under optional-horizon average-run-length control: for every \(\eta>0\), they are equivalent to an alarm family \((T_\gamma)_{\gamma\ge1}\) satisfying \(\Prob_{P^\infty}(T_\gamma\le\sigma)\le \E_{P^\infty}\sigma/\gamma\) for every null law and every stopping time \(\sigma\). In fact, when these conditions hold, we can construct a single e-detector such that every null-law average run length lies between \(\gamma\) and \((1+\eta)\gamma+1\), and having robust Lorden delay \(O_Q(\log\gamma)\).
Read the full abstract ▾
Discrete Mathematics · Cross-listed from math.PR
Optimal mixing of the systematic scan dynamics via approximate tensorization of entropy
Antonio Blanca and Md Tahmidur Rafid
We study the mixing time of the systematic scan dynamics for high-dimensional discrete distributions. This Markov chain updates coordinates sequentially according to a fixed predetermined order, in contrast to the Glauber dynamics that updates coordinates selected uniformly at random. The systematic scan is often favored in practice because it exhibits strong empirical performance, but its theoretical analysis remains far less developed than that of Glauber dynamics. We take a step toward addressing this imbalance by showing that two standard functional notions of weak dependence between the coordinates of the distribution provide strong convergence guarantees for the systematic scan dynamics. First, we show that approximate tensorization of
entropy implies optimal $O(\log n)$ mixing time for every scan order under standard marginal, connectivity, and bounded interaction degree assumptions about the distribution. Second, we show that approximate tensorization of variance yields a constant-factor contraction of the variance functional per scan, which in turn implies an optimal $O(1)$ relaxation time for the natural additive and multiplicative reversibilizations of the systematic scan dynamics. Compared with our
entropy result, the variance bound improves the dependence on the maximum interaction degree from exponential to quadratic and requires weaker assumptions on the distribution. As concrete applications of our results, we establish optimal $O(\log n)$ mixing of the systematic scan dynamics for bounded-degree
antiferromagnetic two-
spin systems in the tree-uniqueness region and for the
ferromagnetic $q$-state Potts model on square boxes in $\mathbb{Z}^2$ throughout its subcritical regime.
Read the full abstract ▾
Artificial Intelligence
The Normalization of Deviance in AI Development
Emilio Barkett, Alexander Kimpton, Daniel Graham et al. (+1 more)
Work on the risks of artificial intelligence has focused predominantly on capability risk: the danger that systems become too powerful, too autonomous, or too misaligned with human values. Far less
attention has been paid to the organizational level—to whether the institutions building these systems are themselves predisposed to drift toward failure. This paper argues that they are. Regardless of how capable AI systems become, the organizations building them face the same structural dynamics that preceded past major technological disasters. Drawing on case studies of the Space Shuttle Challenger, the Three Mile Island accident, and the Boeing 737 MAX crashes, this paper identifies the common structural mechanisms preceding each failure and maps them onto contemporary AI development. The findings suggest that existing safety infrastructure may provide less protection than it appears, as organizations can complete safety processes in full compliance and still produce catastrophic outcomes. The pre-disaster period of AI development is still underway; the purpose of this paper is to make these dynamics legible while they can still be interrupted.
Read the full abstract ▾
All authors Emilio Barkett, Alexander Kimpton, Daniel Graham, Yusuf Kundgol
Machine Learning
A Multi-Source Ensemble Approach to Candidate Generation for Alternative Vacation Rental Property Recommendations
Syed Mohammed Arshad Zaidi, Eric Rincon and Shayan Hassantabar
Alternative property recommendations play a critical role in vacation rental marketplaces, helping users discover relevant options when viewing a specific listing. However, generating high-quality candidate alternatives presents unique challenges: heterogeneous inventory, geographic constraints, rapid availability changes, and long-tail property distributions. We present a comprehensive study of candidate generation (CG) approaches for vacation rental alternatives, comparing collaborative filtering, shallow
embeddings, and graph neural network (GNN) methods. Our experiments on a large-scale vacation rental platform (over 2M active properties) show that a hybrid architecture combining item-based collaborative filtering with GNN-based
retrieval improves Recall@300 by 14.8% over the strongest baseline, by leveraging the complementary strengths of the two sources: collaborative filtering excels at early recall for properties with rich interaction history, while GNNs discover diverse, non-obvious alternatives and handle cold-start scenarios more effectively. As a component result, GNN-based
embeddings alone substantially outperform shallow Hotel2Vec
embeddings (48-68% relative recall improvement across K), motivating their inclusion in the ensemble. Crucially, we examine how CG-stage gains carry through to the downstream ranking stage, and find that a stronger candidate pool yields higher downstream ranking quality, though attributing this effect cleanly is complicated by the coupling between candidate generation and ranker training. This recall-conversion gap is an important consideration for practitioners deploying new
retrieval methods in two-stage recommendation systems.
Read the full abstract ▾
Computer Vision
Gaussian Linear Functional Manifold Method for Massive Point Cloud Data
Hong Zhao, Tonglin Zhang, Baijian Yang et al. (+2 more)
Reconstructing continuous terrain
manifolds from massive, unstructured airborne
LiDAR point clouds remains challenging in complex Wildland-Urban Interface (WUI) environments, where deep neural networks require costly point-wise annotations and nonparametric surface reconstruction methods often lack structural interpretability. This paper introduces the Gaussian Linear Functional
Manifold (GLFM), a physics-informed statistical framework that represents continuous surface topography using deterministic linear functional bases while modeling microscale diffuse
laser backscatter as an isotropic Gaussian process. To avoid the quadratic computational cost of exact constrained maximum likelihood estimation, we develop an algebraic singular value decomposition (SVD) rank-reduction
algorithm that enables linear-time parameter estimation and closed-form quadric classification. Evaluated on 35.2 km^2 of real-world aerial
LiDAR data, GLFM automatically filters ground points and extracts morphological features, achieving an adjusted Rand index (ARI) of 0.9933 against field-verified ground truth and outperforming four leading baselines while maintaining an out-of-core
memory footprint. The framework provides a rigorous, interpretable, and scalable foundation for large-scale point cloud analytics.
Read the full abstract ▾
All authors Hong Zhao, Tonglin Zhang, Baijian Yang, Jin Wei-Kocsis, Songlin Fei
Computation and Language
Some Tokens Behave like Magnets: Revealing Linguistic Organization in the Layers of Language Models
Andrew Liu, Devan Srinivasan and Gerald Penn
We identify a special group of
token vectors inside large
language models (LLMs), which we term
magnetic vectors, that organize the surrounding
tokens by either attracting or repelling them. Particularly,
tokens pointing the same way as an attracting magnet are elongated;
tokens pointing the same way as a repelling magnet are compressed. Just as physical magnets pull or push away the iron filings around them, these vectors organize their surroundings through two opposing polarities. Moreover, we identify a statistically significant pattern in linguistic category where function words consistently act as repelling magnets in early layers, and we also find magnets consistently reorganize their polarities in unique ways deeper in the model. In a further case study we find this observation may unveil a deliberate, layer-wise organization in how LLMs process
language. This pattern is consistent across different LLM architectures, sizes, and layer configurations. It is also causally relevant. When the LLM is fine-tuned for a downstream task, the task-functional
tokens emerge as magnets. E.g., in question answering, the answer-span
tokens become uniquely repelling magnets in the final layer, geometrically carving the answer out of the surrounding context. Furthermore, removing early-layer repelling magnets devastates syntactic tasks (POS tagging accuracy drops from 91% to below 10%) while sparing semantic ones, and removing late-layer attracting magnets does the reverse. We believe this phenomenon warrants further investigation, as it opens the first probe-free path to understanding how
language models geometrically organize linguistic computation across their layers.
Read the full abstract ▾
Computer Vision
SeRV: Semantic-Aligned Residual Vector Quantization for American Sign Language Generation
Hongyu Wu, Xu Wu, Tianhao Wu et al. (+4 more)
American Sign
Language (ASL) generation remains challenging due to limited paired text-ASL motion data and the difficulty of learning motion representations both precise for reconstruction and predictable from linguistic input. Existing methods rely on motion tokenizers optimized for reconstruction, without explicit semantic supervision from paired text. As a result, the learned
tokens remain limited in supporting semantically consistent and fine-grained ASL motion generation. To address this limitation, we propose SeRV (Semantic-Aligned Residual Vector Quantization), a semantic-aligned RVQ tokenizer for ASL generation. SeRV learns a semantically structured residual
token space by combining global-level motion-text
alignment with
token-level text-conditioned supervision. Building on this tokenizer, a Hierarchical GPT predicts residual motion
tokens in a coarse-to-fine manner, generating structurally coherent and semantically aligned 3D ASL motion. We further construct a large-scale reconstructed 3D ASL motion-text
benchmark by recovering paired 3D motion from OpenASL
videos. Experiments on How2Sign and OpenASL show that SeRV outperforms six state-of-the-art baselines, with user studies assessing reconstruction quality and semantic fidelity. Samples of our reconstructed 3D ASL motion are available on our project page: https://signvideos.github.io/Demo/.
Read the full abstract ▾
All authors Hongyu Wu, Xu Wu, Tianhao Wu, Jiawei Yu, Phuc Nguyen, Jian Liu, Yi Wu
Systems and Control
Practical Zero-Trust for Mission-Critical Robotic Fleets via Hardware Attestation and Packet Timing Watermarking
Ryne Gonzales, Ethan Liesdyanto, Rex Worley et al. (+3 more)
Autonomous unmanned vehicles are vital to tactical missions, mission-critical public-safety operations like search and rescue and disaster response. However, their reliance on open wireless links and standard
Robot Operating System (ROS 2) middleware exposes a broad cyber-physical attack surface. A compromise of these systems can disrupt real-time control loops, leading to mission failure or asset loss in high-stakes environments. This paper presents and empirically evaluates a layered, context-aware cybersecurity framework enforcing Zero-Trust principles for a multi-node robotic fleet over Wi-Fi. The framework integrates an active hardware root of trust (TPM 2.0), centralized in-band and out-of-band SIEM telemetry monitoring (ELK Stack and Kismet), and a non-cryptographic Inter-Packet Delay (IPD) timing watermark. Evaluated on a live ROS 2 mobile testbed under multi-layer exploits (OSI Layers 2-5), results demonstrate that while volume-based filters isolate brute denial-of-service floods, tracking the statistical sample kurtosis (K) of the embedded IPD watermark exposes stealthy Man-in-the-Middle command injections with complete detection accuracy without payload overheads.
Read the full abstract ▾
All authors Ryne Gonzales, Ethan Liesdyanto, Rex Worley, Michael Frederick, Jaewon Kim, Eman Hammad
Multiagent Systems
SimTIO: A Simulation-Grounded Multi-Agent LLM Framework for Compositional Traffic Intervention Optimization
Shuyang Li and Ruimin Ke
Traffic analysts must translate diagnosed bottlenecks into executable interventions without allowing local improvements to degrade network-wide performance. This study presents SimTIO, a simulation-grounded multi-
agent large
language model framework for composing and selecting traffic interventions under explicit operational constraints. SimTIO first simulates an unmodified SUMO scenario to identify a baseline-frozen set of ten bottleneck edges. A grounded sampler then initializes signal-control, corridor-speed, and demand-preserving routing actions, while three specialist
agents use measured simulation feedback to select one-parameter refinements from validator-confirmed mutation catalogs. Compatible actions are combined and re-simulated so that their interaction effects are measured rather than inferred. Final selection minimizes bottleneck time loss while constraining network-wide delay, neighboring-road spillover, throughput loss, and teleport events, with the unmodified scenario retained as a no-operation guard. Across 15 cases covering five U.S. urban networks, three synthetic-demand seeds, and 2,400 origin-destination trips per scenario, SimTIO reduced Top-10 bottleneck time loss by an average of 9.18 percent and network-wide delay by 2.78 percent. It found a feasible improving plan in 86.7 percent of cases, compared with 73.3 percent for grounded random search and 80.0 percent for a deterministic heuristic under the same seven-simulation budget, although the differences in Top-10 improvement were not statistically significant. These results support using LLMs as constrained, feedback-guided local search operators while reserving final decision authority for executable tools, microscopic simulation, and explicit safety constraints.
Read the full abstract ▾
Systems and Control
Resilient Control Loops in Autonomous Vehicles Under Adversarial Jamming via Spectral Perception and Network-Layer Failover
Luis Barajas, Colin Jeardoe, Jaewon Kim et al. (+1 more)
The operational integrity of autonomous mobile
robots relies on the continuous availability of wireless control loops, making them highly attractive targets for
adversarial intentional electromagnetic interference. This paper introduces a resilient, cross-layer framework that combines physical-layer spectral perception with network-layer routing optimization to protect middleware stability, such as ROS2, during intentional electromagnetic interference. Utilizing a software-defined
radio front-end, the system extracts dynamic spectral descriptors, including spectral
entropy and channel occupancy, to inform a Random Forest classifier that establishes adaptive environmental baselines. To ensure uninterrupted data flow, the architecture maintains dual pre-authenticated physical interfaces in a hot-standby configuration, enabling instantaneous failover through automated network routing table updates. Empirical validation on a physical ROS2 mobile
robot testbed demonstrates that this adaptive hardware-assisted architecture optimizes communication recovery to an average of 141ms. This sub-second restoration translates directly into a 78.9% reduction in pooled root-mean-square path tracking error compared to software re-association, successfully securing system-level mission integrity.
Read the full abstract ▾
All authors Luis Barajas, Colin Jeardoe, Jaewon Kim, Eman Hammad
Computer Vision
RenderFormer-V2: Neural Rendering with Heterogeneous Scene Primitives
Chong Zeng, Yue Dong, Pieter Peers et al. (+2 more)
We present 'RenderFormer-V2', a unified learned
transformer-based neural rendering model, complementary to modern physics-based rendering systems, that can handle diverse light-transport effects such as caustics, volumetric scattering, environment lighting, textured and displaced surfaces and out-of-distribution materials without per-scene training or specialized code. RenderFormer-V2 models global light transport as a sequence-to-sequence transformation. Following its predecessor, RenderFormer-V2 also employs a two stage process: a view-independent stage that resolves intra-scene primitive to primitive transport, and a view-dependent stage that transforms the internal neural scene representation into image pixels. Different from RenderFormer, our model employs a novel combined windowed-
attention and rendering-informed
attention sink in the view-independent stage to improve scalability while maintaining render accuracy. To further improve versatility, RenderFormerV2 supports heterogeneous scene primitives, including environment maps and participating media, and it employs a material encoding independent of the underlying surface reflectance model that encodes material appearance via a novel neural
embedding. We demonstrate the versatility of RenderFormer-V2 on a variety of scenes and perform an extensive ablation of the improved
attention mechanism.
Read the full abstract ▾
All authors Chong Zeng, Yue Dong, Pieter Peers, Lvmin Zhang, Maneesh Agrawala
Emerging Technologies · Cross-listed from quant-ph
SAQC: A SAT-Aware Compilation Framework for QAOA-Based Quantum Optimization
Till Schnittka, Julie Maria Raju, Abhoy Kole et al. (+1 more)
The
Quantum Approximate Optimization
Algorithm (QAOA) is a promising variational approach for solving Boolean satisfiability (SAT) problems. Existing SAT-to-QAOA workflows first translate Boolean formulas into penalty
Hamiltonians before applying generic
quantum compilation, thereby discarding SAT-specific information such as clause structure, logical operators, and variable dependencies. Consequently, subsequent compiler optimizations cannot exploit the underlying Boolean formulation. This paper presents SAQC, a SAT-aware compilation framework that introduces a SAT Intermediate Representation (SIR) to preserve Boolean structure during compilation. The proposed framework supports both CNF and XOR-extended CNF (eCNF) formulations and performs SAT-aware optimizations including clause rewriting, dependency analysis, and clause scheduling prior to
quantum lowering. Building on the optimized SIR, SAQC provides a unified framework for both penalty
Hamiltonian generation and direct clause-to-ansatz synthesis, together with ansatz optimizations based on relative-phase decomposition, ancilla-assisted synthesis, and dynamic uncomputation. Experimental results demonstrate significant reductions in
circuit depth, two-
qubit gate count, and compilation time compared with conventional
Hamiltonian-based workflows while remaining compatible with existing
quantum compilation frameworks.
Read the full abstract ▾
All authors Till Schnittka, Julie Maria Raju, Abhoy Kole, Rolf Drechsler
Artificial Intelligence
Beyond Prompts: Measuring and Optimizing LLM Tool-Agent Harnesses
Cen Mia Zhao, Haibo Ruan, Wenjie Chen et al. (+3 more)
LLM tool
agents can be improved without retraining by modifying the runtime harness around a fixed model:
prompts, tool interfaces, middleware, state handling, and recovery logic. We study this setting as resource-bounded harness selection for fixed-model multi-turn tool
agents, with the search surface scoped to
prompts and tool-boundary middleware: edits are guarded intercepts at the tool boundary, not arbitrary rewriting of
agent execution logic. Our optimizer-agnostic protocol reports mean held-out lift, worst-condition lift, repeatability, logged cost diagnostics, and RelLift95(B), a conservative estimate of the held-out gain of the harness selected under budget B. We instantiate the protocol with
prompt-only and
prompt-plus-middleware optimizers, including PRISM, which clusters failures and routes repairs to
prompt, tool-boundary middleware, or joint edit surfaces within a Pareto search. On BFCL multi-round, tau2-Retail, and tau2-Telecom, PRISM obtains mean held-out lifts of 14.2, 14.9, and 10.1 percentage points and positive empirical RelLift95 on all three
benchmarks, and a component ablation attributes the margin chiefly to failure-surface routing and the edit-pattern constraint. Across optimizers, the results show that some search procedures can occasionally find large gains but still choose brittle updates, so the reliability of the chosen harness should be reported alongside average held-out lift.
Read the full abstract ▾
All authors Cen Mia Zhao, Haibo Ruan, Wenjie Chen, Pei-fen Tu, Usman Abbasi, Joel Hesch