Artificial Intelligence · Cross-listed from cs.LG
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
Artificial Intelligence · Cross-listed from cs.AR
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
Artificial Intelligence · Cross-listed from cs.CV
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 ▾
Artificial Intelligence · Cross-listed from cs.PF
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
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
Artificial Intelligence · Cross-listed from cs.DB
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
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
Artificial Intelligence · Cross-listed from cs.CV
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
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
Artificial Intelligence · Cross-listed from cs.CV
Bigger Text Encoders Can Hurt CLIP Zero-Shot Performance
Samir Char, Carles Domingo-Enrich and Randall Balestriero
Contrastive
Language-Image Pretraining (CLIP) is a building block of many machine learning applications. Scaling laws have guided resource allocation for large-scale training, yet prior work treats total CLIP model size as a single variable, without exploring how the capacity split between encoders impacts downstream performance. Here, we train multiple CLIP models with different
vision and text encoder sizes, revealing that for most
vision encoders, there is an optimal text encoder size beyond which zero-shot performance degrades—even as total parameter count increases. Exploiting this behavior yields efficient configurations that match the zero-shot performance of the standard ViT-B/16 architecture with up to 55% fewer parameters. We further show that this degradation stems from overfitting induced by the oversized text encoder, and that using modality-specific weight decay coefficients not only recovers but improves performance across all degraded configurations. A
geometric analysis reveals a trade-off in which scaling the text encoder improves
embedding uniformity but worsens cross-modal
alignment; we further show that these metrics are predictive of zero-shot performance. We hope these findings motivate CLIP architectures and training methods that counteract this degradation, a prerequisite for scaling CLIP reliably and efficiently.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
Newton Matching for Generative Modeling: A Unified Framework for Fine-Tuning and Sampling
Zeyang Li, Yunan Wang, Paolo Giaretta et al. (+1 more)
We develop Newton Matching, a unified framework for fine-tuning and sampling in generative modeling. The target is $\pi\propto\mu e^{\tau r}$, where $r$ is the reward, $\tau>0$ the inverse temperature, and $\mu$ denotes the pretrained model's terminal density for fine-tuning or the constant $1$ for sampling. We shift the paradigm from isolated losses to iterative optimization over canonical models: population minimizers of standard conditional matching for terminal densities. Under compatible smooth-realization assumptions, canonical velocities form a
manifold diffeomorphic to the density
manifold. Transporting the Fisher-Rao metric and mixture connection to this
manifold, we show that the reverse-KL Hessian equals the metric, so the Newton direction coincides with the negative Fisher-Rao gradient. At terminal density $\rho$, each stage takes a tangential step generated by the regularized reward $r-\frac1\tau\log(\rho/\mu)$, followed by terminal-density-preserving canonicalization. This canonical retraction yields an exact finite-stepsize density characterization. For the ideal iteration, we prove strict reverse-KL descent away from the target for $0 < \eta \le \tau$, global convergence under mild conditions, and local quadratic convergence for full steps ($\eta=\tau$). Covariance and gradient forms, each with forward or reverse regression-pair constructions, yield sample-wise tangential-update losses with the same population minimizer, without importance sampling or full-trajectory backpropagation. We develop approximate updates and define critical-point consistency as vanishing tangential displacement if and only if $\rho=\pi$. We recover representative methods as exact realizations, critical-point-consistent approximations, or objective-altering variants, enabling modular
algorithm design. Our work advances the theory and
algorithms of reinforcement learning for generative models.
Read the full abstract ▾
All authors Zeyang Li, Yunan Wang, Paolo Giaretta, Navid Azizan
Artificial Intelligence
Recovering Temporal and Geographic Signals from Language Model Embeddings
Esteban Feuerstein, Victoria Klimkowski, Juan Manuel Ortiz de Zarate et al. (+1 more)
Understanding whether
language-model
embeddings encode structured real-world information is important for both representation analysis and information
retrieval. We study this question for temporal and geographic signals using a simple projection-based method that operates directly on output
embeddings. Given a small set of seed examples, the method defines an axis in
embedding space and ranks texts or entities by their projection onto that axis. Our approach is fully black-box and model-agnostic: it requires only
embeddings, without access to model weights, internal activations, auxiliary probes, or additional training. This makes it applicable to modern
embedding models available only through APIs and provides a lightweight way to analyze whether temporal and spatial dimensions are present in their representation spaces. We apply the method to temporal and geographic
datasets and find that
embedding projections recover meaningful chronological and spatial structure. These results provide evidence that output
embeddings encode signals relevant to time and space, while also offering a practical tool for interpretability and for downstream temporal and geographic information
retrieval tasks, such as temporal ordering, geographic ranking, and tagging.
Read the full abstract ▾
All authors Esteban Feuerstein, Victoria Klimkowski, Juan Manuel Ortiz de Zarate, Federico Hernán Suaiter
Artificial Intelligence
CUSP: Decomposable Collective Uncertainty for Multi-Agent Multimodal Reasoning
Chung-En Johnny Yu, David Garcia, Brian Jalaian et al. (+1 more)
Aggregating heterogeneous
vision-
language models (VLMs) can improve
multimodal reasoning, but neither an individual model's confidence nor that of the aggregated answer measures reliability at the system level. We present CUSP (Collective Uncertainty through Semantic Opinion Pooling), a training-free uncertainty quantification framework that maps multiple VLM responses to a shared semantic response space, pools them into a pooled semantic opinion, and reports two complementary system-level signals: collective uncertainty, the dispersion of the pooled opinion, and Jensen-Shannon divergence (JSD), the conflict among the model-level opinions. Within this pooled semantic opinion, the unnormalized collective
entropy decomposes exactly into the mean of the models' individual semantic
entropies and the JSD, separating total dispersion from model conflict. Requiring neither
token logits nor calibration labels, CUSP applies to open-weight and commercial VLMs alike. In static multi-VLM ensembles, collective uncertainty is the strongest signal in the small-model regime (0.764 AUROC for prediction-error detection, 0.889 AUARC for abstention), outperforming uncertainty baselines majority voting and naive selection by 4.7 to 15.8 points and widening its margin as the ensemble grows; JSD is strongest in the evaluated commercial regime (0.819 AUROC, 0.910 AUARC) and ranks hard-answer model conflict with AUROC up to 0.982. The pooled prediction also improves accuracy over the average single model by 5.6 to 13.0 points. Over the full trajectory of a multi-step, multi-
agent system, subagent collective uncertainty ranks system failures above chance (0.619 AUROC) and gives the best abstention ordering among the evaluated signals (0.699 AUARC).
Read the full abstract ▾
All authors Chung-En Johnny Yu, David Garcia, Brian Jalaian, Nathaniel D. Bastian
Artificial Intelligence · Cross-listed from cs.CR
XAI-SDN: An Explainable Entropy-Guided Machine Learning Framework for Real-Time DDoS Detection in Software Defined Networks
Adeel Ahmad, Ali Akarma, Ahmad Ali et al. (+2 more)
One of the biggest risks faced by Software Defined Networks (SDN) is the Distributed Denial of Service (DDoS) attack in which a compromised controller can make an entire network unusable. To address these challenges, we suggest an
entropy-guided machine learning framework, called XAI-SDN, for real-time DDoS detection in SDN environments which is lightweight and explainable. The framework extends the flow features extracted by CICFlowMeter with eight Shannon
entropy metrics obtained by an $\mathcal{O}(1)$ rolling
algorithm and uses a Random Forest classifier with SHAP TreeExplainer for providing transparency at the prediction level. On a fixed temporal split, XAI-SDN achieves an accuracy of 99.9987%, a macro F1-score of 99.9621%, and an AUC-ROC of 1.0000 on the full 3.59 million flows of the CIC-DDoS2019 SYN
benchmark. The pipeline sustains 0.0165~ms per flow (60{,}606 flows/s) without the use of SHAP and 0.5122~ms per flow (1{,}953 flows/s) with full support of SHAP under the 99.14% prevalence of DDoS traffic, which is a step towards achieving a balance between the detection performance and operational transparency in next-generation SDN security.
Read the full abstract ▾
All authors Adeel Ahmad, Ali Akarma, Ahmad Ali, Hammad Muneer, Toqeer Ali Syed
Artificial Intelligence · Cross-listed from cs.LG
Analysis of Respiratory Sinus Arrhythmia with Neural Networks
Julian Szymanski, Patryk Orkisz and Higinio Mora
The paper introduces a neural network-based approach for analyzing ECG signals to estimate respiratory rate by leveraging the phe- nomenon of Respiratory Sinus Arrhythmia (RSA). Our method employs a deep learning model trained to predict respiratory waveforms directly from ECG input data. To achieve this, we developed and evaluated three different neural network architectures capable of automatically extract- ing relevant features from ECG signals without the need for manual preprocessing. The proposed approach offers a robust and scalable solu- tion for non-invasive respiratory monitoring, with potential applications in healthcare and wearable technology
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.CL
Who Maintains Agent Skills? A Longitudinal Study of Human-Governed, AI-Assisted Skill Maintenance
Chen Shen and Estevam Hruschka
Lifelong LLM
agents increasingly rely on external skill artifacts as one element for preserving and reusing capabilities over time. These skills (usually portable Markdown files such as SKILL.md) describe when and how to apply a capability and must be corrected, expanded, and consolidated as tools and usage patterns shift over deployment. Recent work seeks to automate skill curation, but it largely evaluates against automated baselines and treats human maintenance as an unmeasured bottleneck. We study that missing process directly. We mine the full commit histories of five public AI-skill repositories, a purposive sample of AI-tooling organizations, covering 873 commits, 143 skill files, and 254 substantive post-creation edits from October 2025 to June 2026. We code each edit with pre-registered governance, operation, and trigger-evidence codebooks. Three findings emerge. First, every substantive edit is authored or merged through a named human account, while 62% carry an AI co-author trailer, with large repository-level variation. Second, these edits are genuine curation: an audited sample shows that most change skill content, and the coded operations are dominated by additions and corrections. Third, a pre-registered rule-likeness axis fails its reliability gate; reliably coding rule-likeness from commit artifacts remains an open measurement problem. We release the corpus, codebooks, mining scripts, and a replay protocol for automated skill curators. For self-evolving
agents, public skill maintenance currently looks less like an autonomous pipeline than a human-governed, AI-assisted loop that future curators must measure against and operate within.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
PAC-Private Autoregressive Generation: Calibrating Noise to Ensemble Disagreement
Mina Mirzadehsarcheshmeh and Amir Keyvan Khandani
Language models adapted on private text are often served through APIs, so
privacy leakage occurs through generated outputs rather than exposed weights. Private prediction protects these releases. Methods such as PMixED incur
privacy cost at each release and increasingly rely on the public model over long horizons. PAC
privacy instead calibrates noise to output variability across possible secrets, adding less noise when predictions are stable. To our knowledge, PAC-private prediction has not previously been extended from classification to autoregressive generation. We construct $m=128$ overlapping worlds from the private corpus, with each record appearing in exactly $m/2$ worlds, and train one adapter per world over a frozen public model. The realized world is the secret. At each
token, the public model defines a candidate set, the worlds vote, and their posterior-weighted disagreement determines the PAC noise; unanimity requires no calibration noise. We prove $I(S;Y_{1:T}) \leq I(S;H_T) \leq bT$. Our contributions are extending PAC
privacy to autoregressive generation, handling adaptive self-generated contexts, and introducing coupled decoding that preserves
privacy accounting while avoiding greedy degeneration. On WikiText-103 with GPT-2-small, we retain 74% of the fine-tuning gain at a per-
token budget of $2^{-32}$, while membership-inference success is bounded by 51.08% after $10^6$
tokens; posterior-
entropy estimates of leakage are roughly 17% of the charged budget. Inference
privacy is not content protection: even when membership advantage on a memorized canary is indistinguishable from zero, the canary is emitted at the same rate. Against PMixED under matched membership-inference bounds on the same data universe and test set, we retain 98% of non-private headroom from $10^2$ to $10^6$
tokens, versus at most 56%, with no crossover.
Read the full abstract ▾
Artificial Intelligence
What LLM Trading Agents Actually Do in Production: A Six-Month, Population-Scale Record from Two Fleets
T.J. Barton, Chris Constantakis, Patti Hauseman et al. (+4 more)
We present a continuous, population-scale measurement record of autonomous
language-model trading
agents operating in production across two systems with one design lineage: DX Terminal Pro (3,505 user-funded vaults trading real ETH in Base memecoin
markets for 21 days, February to March 2026) and the DXAP live alpha fleet (500 to 599 user-created
agents all-history, 91 to 117 concurrently active, trading Hyperliquid perpetuals, June to August 2026). The record spans roughly six months, 7.5M single-model invocations with about 300K onchain actions, and a further 231,638 multi-tool turns producing 14,596 fills. Four findings carry the paper. First, the operating layer determines behavior more than anything written in strategy text: a risk slider explains leverage (+0.425 per level),
agent fixed effects absorb 60% of variance, and a leaderboard render boundary causally routes selection (regression discontinuity 1.75x at the top-3 cut). Second, sizing is
volatility-blind: median leverage is 5.0x in every
volatility sextile, and one posture-slider
cell (11% of the book) holds 62% of liquidations. Third,
agents capture almost none of the upside they reach: 43.2% of positions saw at least +300 bps of favorable excursion within 24h, yet 49.3% of those closed with a negative trade return; a mechanical bracket recovers +39.0 bps per position. Fourth, neither fleet shows a directional edge. The DXAP fleet is not profitable and trails a matched Hyperliquid retail
benchmark (41% vs. 50% roundtrip win rate). A paired-replay league of frontier models on 416 captured production scenarios finds decision quality statistically indistinguishable at this horizon, while choice stability differs sharply across model families. Every headline survives day-clustered inference, permutation nulls, and a common-fee restatement; the paper closes with a 17-rule methodology canon bought with our own retractions.
Read the full abstract ▾
All authors T.J. Barton, Chris Constantakis, Patti Hauseman, Annie Mous, Alaska Hoffman, Brian Bergeron, Hunter Goodreau
Artificial Intelligence · Cross-listed from cs.CV
Full-Page Optical Music Recognition of Handwritten Monophonic Scores
Adrian Rosello, Antonio Ríos-Vila, David Rizo et al. (+1 more)
Full-page end-to-end Optical Music Recognition seeks to transcribe entire music pages directly into symbolic notation, avoiding the limitations of traditional pipelines that rely on accurate staff
segmentation. Recent
Transformer-based architectures have achieved strong performance on typeset scores, relying on large-scale synthetic data for pretraining. However, their applicability to handwritten music remains largely unexplored. In this work, we study full-page transcription on handwritten monophonic collections and analyze the impact of synthetic pretraining in this setting. To investigate which factors are most relevant during pretraining, we introduce a generator capable of producing visually coherent full-page scores in both typeset and handwritten styles. Experiments on three real handwritten
datasets provide a comparative evaluation of several full-page pipelines and different synthetic pretraining strategies. The results suggest that the benefits of synthetic pretraining are primarily associated with learning structural layout conventions rather than with visual similarity to the target handwriting.
Read the full abstract ▾
All authors Adrian Rosello, Antonio Ríos-Vila, David Rizo, Jorge Calvo-Zaragoza
Artificial Intelligence
The convergent laboratory: when AI reasoning, autonomous experiments, high performance and quantum computing reshape chemistry
Eliu Huerta, Xiaoyun Wang, Geetika Gupta et al. (+7 more)
This Comment emerges from TPC26 (https://tpc26.org), a conference convening leaders from academia, national laboratories, and industry who are reshaping materials science discovery. The meeting explored how AI, autonomous
agents, self-driving labs, higher performance and
quantum computing converge to amplify their individual impact on materials science discovery. The perspectives here reflect the firsthand experiences of researchers at these frontiers and capture the essence of this global endeavor. As AI-driven
reasoning, autonomous agentic frameworks, self-driving laboratories, and fault-tolerant
quantum processors mature simultaneously, we offer this Comment as a reference at what we believe is a tipping point of transformative advances and productive disruption in the chemical sciences.
Read the full abstract ▾
All authors Eliu Huerta, Xiaoyun Wang, Geetika Gupta, Edward H. Sargent, Cameron J. Owen, Victor Fung, Abhishek Mitra, Austin Cheng, Emma Bouchard, Shams Mehdi
Artificial Intelligence · Cross-listed from cs.CL
TamilEOT: A Dataset and Model for Semantic End-of-Turn Detection in Tamil Telephone Speech
Santhoshkumar V
A voice
agent has to decide, at every pause, whether the user has finished speaking. Without a model of the
language that decision falls back to a fixed silence timeout: set it short and the
agent interrupts, set it long and every turn pays the full wait. Open semantic end-of-turn detectors exist, but to our knowledge none covers a South Indian
language. We release TamilEOT: 18,485 labelled turn boundaries cut from 116 real Tamil telephone conversations, and two audio-only detectors fine-tuned from Smart Turn v3. On a held-out split of 4,168 clips from 30 unseen calls, accuracy rises from 70.30% zero-shot to 83.71% (8.7 MB) and 86.13% (21 MB); ROC-AUC rises from 0.751 to 0.921. Both models run in under 150 ms single-threaded on a laptop CPU. We also report what building it cost. Rule-derived labels, checked against a blind human listening pass, were right 95.9% of the time on the positive class and 44.4% on the negative class, which is below chance, because the rule answered a different question than the model is asked. Replacing them with an audio-LLM labeller measured at 97.5% human agreement cost US$5.69. Of every training lever we measured, only encoder capacity moved the result; three runs at identical config and seed span 0.87 accuracy points, which is the floor below which none of our other deltas mean anything. Replaying the same labelled boundaries through the production VAD and streaming adapter costs a further 2.60 points, and 7.8% of boundaries are never surfaced to the model at all. Data, weights, code and every negative result are public.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.RO
Adaptive Cost-Sensitive Machine Learning for Autonomous Robot Navigation Failure Prediction: When Not All Errors Are Equal
Rifa Ferzana
Autonomous
robot navigation failures differ not only in categorical severity but also in the physical context in which they occur. A near-miss at low speed under reliable sensing is not equivalent to the same event during rapid motion, close obstacle approach or degraded perception. This paper reframes
navigation failure prediction as consequence-sensitive
forecasting. We first establish a fixed baseline in which training weights are modulated by categorical severity, then introduce an adaptive extension defining a state-dependent consequence function combining severity with normalised velocity, obstacle proximity and sensing uncertainty, together with a risk-sensitivity term that rises as conditions deteriorate. We evaluate on 2,000 simulated differential-drive episodes (~1,000,000 timesteps) using episode-level GroupKFold, with external validation on the UCI SCITOS G5
dataset. Fixed weighting raises Logistic Regression high-severity recall from 0.851 to 0.985 and reduces missed consequence cost from 1,940 to 313; the adaptive extension reaches 0.998 and 82. Under matched false-positive conditions, however, the discriminative advantage is modest (0.986 versus 0.984), so most of the gain reflects a more conservative operating point rather than better ranking. The effect is consistent across all five folds and stable across a threefold span of context coefficients. Because the primary simulation produced no collisions, we add a controlled extension in which 108 of 600 episodes terminate in contact: collision recall rises from 0.850 to 0.966 (fixed) and 0.984 (adaptive), with missed collision cost falling from 1,000 to 105, at false-positive rates of 0.413 and 0.799, respectively. Context-dependent consequence modelling thus provides a principled mechanism for allocating conservatism by physical risk.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from physics.soc-ph
WolfSociety: Understanding Collective Risk from Harmful-Agent Scaling in Financial Agent Societies
Lejun Zhang, Sarah Lu-Liang, Xin Jiang et al. (+3 more)
Safety evaluations typically focus on individual
agents, but interacting
agents can spread harmful information and influence the environment in which later decisions are made. We study how collective failure changes with harmful-
agent fraction and society size in a controlled financial
agent society, where
agents communicate over a social network and trade in a shared
market. In the primary financial scenario, collective failure requires broad harmful
diffusion together with severe price dislocation or liquidity stress. Across all tested society sizes, failure remains rare at low harmful fractions but rises sharply over a narrow range. As society size grows from N=100 to N=2000, the harmful fraction associated with a 50% failure probability decreases from 4.7% to 2.2%, while the corresponding number of harmful
agents increases from approximately 5 to 44. In contrast, when the number of harmful
agents is held fixed, their impact becomes weaker as the society grows. Controlled interventions further show that broader network reach shifts the collapse boundary toward lower harmful fractions, whereas stronger conformity alone has little effect. To characterize these effects, we introduce
Agent Society Dynamics, a finite-size framework for relating harmful-
agent fraction, society size, and interaction structure to collective failure. Overall, our results reveal a nonlinear, size-dependent collapse transition in financial
agent societies, showing that collective failure depends not only on the prevalence of harmful
agents but also on the size and interaction structure of the surrounding society. Code is available at https://github.com/SAIL-Research-Lab/WolfSociety.
Read the full abstract ▾
All authors Lejun Zhang, Sarah Lu-Liang, Xin Jiang, Muning Wen, Weinan Zhang, Shangding Gu
Artificial Intelligence
Agents' Overreliance on Unreliable Tools
Hoyeol Yang, Woojung Song, Taewon Kim et al. (+3 more)
LLM
agents use tools to access information and perform computations beyond their parametric knowledge. Existing tool-use
benchmarks evaluate whether
agents select and call the right tools, assuming that tool returns are reliable. However, tools can return plausible but incorrect outputs. We evaluate 14 models with three tools (web search, an LLM sub-
agent, and a code executor), corrupting their returns to examine whether
agents overrely on unreliable tools.
Agents adopt corrupted returns at high rates, with mean adoption exceeding one third for every tool and reaching 68.0% for web search. Corrupted returns also frequently override correct answers that
agents give without tools, and this reliance persists in more complex tasks and in tasks that combine multiple tools.
Agents often make more tool calls under corrupted returns, and
reasoning traces show them questioning a return and even stating the correct answer. Nevertheless, they still pass the corrupted content to users, rarely warning them of the conflict. We further test interventions spanning
prompts, tool metadata, post-training, and activation steering. Verification
prompts reduce adoption across all three tools while largely preserving accuracy with correct returns, and reliability labels provide partial mitigation in web search. The tested post-training methods offer limited gains, and the effect of activation steering depends on the model and intervention strength. Overall, our findings provide a basis for diagnosing tool overreliance and for developing
agents that assess tool returns rather than assuming their reliability.
Read the full abstract ▾
All authors Hoyeol Yang, Woojung Song, Taewon Kim, Jonghyun Song, Seoyeon Park, Yohan Jo
Artificial Intelligence · Cross-listed from cs.CV
ViT3Flow: A Test-Time Training Transformer MeanFlow for Postoperative Radiograph Synthesis in Scoliosis
Rui Tang, Sicheng Yang, Moxin Zhao et al. (+6 more)
Predicting postoperative spinal morphology from preoperative radiographs could provide valuable support for scoliosis
surgical planning, but remains challenging because
surgical correction induces large spatial changes while anatomical structures must be faithfully retained. We formulate this problem as postoperative scoliosis radiograph synthesis and construct ScoliSurg, the first paired
dataset for this task, comprising 632 preoperative–postoperative whole-spine radiograph pairs with structured morphology information. We further propose ViT$^{3}$Flow, a single-NFE conditional MeanFlow framework for efficient postoperative radiograph synthesis. ViT$^{3}$Flow models
surgical correction as finite-interval generative transport and replaces conventional self-
attention with test-time-training
token mixers that perform sample-specific inner adaptation to the anatomy and deformity pattern of each case. In addition, a Spinal Morphology Extraction
Agent extracts distributions of dominant-curve region and direction from the preoperative radiograph. These distributions guide Diagnosis-Routed Interval Cross-
Attention (DRICA), which performs interval-dependent vertical, horizontal, joint, and global
retrieval from a separate preoperative
token stream. This design enables the evolving postoperative representation to incorporate spatially corresponding anatomical evidence throughout the transport process. Extensive experiments on ScoliSurg demonstrate that ViT$^{3}$Flow achieves the best performance among the compared methods in perceptual image quality, anatomical fidelity, and clinically relevant
geometric accuracy, while requiring only a single network evaluation. These results highlight the potential of ViT$^{3}$Flow for efficient and anatomically faithful postoperative radiograph synthesis in scoliosis
surgical planning.
Read the full abstract ▾
All authors Rui Tang, Sicheng Yang, Moxin Zhao, Hongqiu Wang, Guankun Wang, Lei Zhu, Hongliang Ren, Menglin Cong, Nan Meng
Artificial Intelligence
Planning and Scheduling Business Processes under Control-Flow Uncertainty: Extended Version
Michel Kunkler and Stefanie Rinderle-Ma
Scheduling activities in business processes can improve efficiency (e.g., reduce makespan), but is challenging because the exact sequence of activities required to complete a case is often uncertain due to decisions based on data that emerges during execution. Nevertheless, probabilistic information regarding such decisions can often be estimated or derived from historical execution logs, and can help anticipate which execution paths are likely to lead to successful completion. Planning with particular execution paths affects feasibility, i.e., the probability of successful completion, and the expected number of superfluous activities that are planned but never executed. We frame the problem as a chance-constrained optimization problem and present two formulations: A decomposed approach with two stages, a planning stage that minimizes the expected number of superfluous activities subject to a feasibility constraint, and a scheduling stage that minimizes the makespan over the planned activities; and an integrated approach that combines planning and scheduling into a single formulation. Evaluation on two real-world and one synthetic
dataset shows that the integrated approach yields superior makespans but is intractable at scale, while the decomposed approach scales to large settings.
Read the full abstract ▾
Artificial Intelligence
EnvCraft: Synthesizing Executable Environments in Agentic RL for Claw-like Agent
Yirong Zeng, Shen You, Jinhang Feng et al. (+8 more)
The paradigm of LLMs has rapidly shifted from passive
language interfaces to autonomous Claw-like
agents that execute long-horizon tasks across stateful workspaces. While Agentic Reinforcement Learning (Agentic RL) provides a promising path to optimize these
agents, its scaling is heavily bottlenecked by the severe scarcity of interactive training environments. Existing synthetic environments are strictly limited to tool-calling endpoints, rendering them insufficient for accommodating the end-to-end real-world demands of claw-like
agents. To bridge this gap, we introduce EnvCraft, an automated framework for synthesizing executable environments and scalable training data. Specifically, EnvCraft employs an environment synthesis engine to build sandbox-isolated workspaces, alongside a
topology-aware data generation engine to produce coherent task trajectories. Overall, we synthesize 139 interactive environments comprising approximately 20K complex tasks for Agentic RL training. Experiments on Qwen3/3.5 models (8B-32B) show that our method yields gains of up to +11.9% on Claw-style
benchmarks and +8.0% on general tool-use
benchmarks, with concurrent reductions in inference
token cost. The results confirm that synthesized executable environments provide robust and generalizable learning signals for training.
Read the full abstract ▾
All authors Yirong Zeng, Shen You, Jinhang Feng, Yufei Liu, Xiao Ding, Yutai Hou, Hao Cong, Yuxian Wang, Wu Ning, Wang Xu, Bibo Cai
Artificial Intelligence
Deep belief networks are exact
Gleb Smirnov
We prove that every strictly positive probability distribution on \(\{-1,1\}^n\) is represented exactly by a sigmoid belief network with finite parameters. This answers a question of Sutskever and Hinton. The proof upgrades their probability-sharing approximation to exact representation using Brouwer's fixed-point
theorem.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.HC
Diffusion TV: Experiencing Diffusion Models through Tangible, Embodied Interaction
Sihwa Park
Diffusion TV is an interactive AI art installation that offers a tangible and
embodied experience of
diffusion models through a modified CRT TV. By physically manipulating the TV's
antenna, audiences control the clarity of AI-generated images and sounds, metaphorically enacting the denoising process that underlies
diffusion-based generation. Using the tuning knob, participants switch between three channels featuring AI-generated animals from the Past (extinct species), Present (endangered species), and Future (speculative creatures), situating the interaction within a temporal and ecological narrative. Through continuous audiovisual feedback and physical interaction,
Diffusion TV foregrounds the generative process over final outputs, allowing audiences to explore intermediate states as experiential material. Rather than providing explicit technical explanation, the work presents an alternative,
embodied mode of explainable AI that invites exploratory engagement with and reflection on generative technologies.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
RegionFed: Federated Learning for Personalized Query Understanding in Heterogeneous Retail Environments
Quoc H. Nguyen, Ali Lafzi, Abhijeet Phatak et al. (+4 more)
Retail search systems serve diverse geographic regions with distinct query patterns, vocabularies, and product preferences, creating significant data heterogeneity that challenges both
privacy-preserving training and model personalization.
Federated learning offers a natural solution for
privacy, but standard FL methods produce global models that sacrifice regional performance, while existing personalized FL approaches operate at the parameter level and catastrophically collapse on modern
transformers (below 10% accuracy on T5) due to tied
embeddings and LayerNorm interactions. We introduce RegionFed, an architecture-robust
federated learning framework that sidesteps this failure by operating entirely at the gradient level. RegionFed uses the $\ell_2$ conflict between regional and global gradients as a unified signal that (i) diagnoses heterogeneity, (ii) routes each region to the cheapest sufficient personalization strategy, and (iii) adaptively controls personalization strength. Because it treats models as differentiable black boxes, RegionFed deploys on T5-Small, T5-3B, RoBERTa, and CNN with zero code changes, providing large gains on
transformers (where parameter-level methods collapse) and consistent improvements on CNNs. Across three public
datasets (Amazon ESCI, Amazon Reviews, LEAF-FEMNIST) and four architectures, RegionFed-Meta achieves 92.27%, closing the gap to the
privacy-violating centralized upper bound (Centralized + Regional Weighting: 92.04%, $\Delta$=0.23pp, within 1$\sigma$) while providing $(\epsilon{\approx}0.60)$-differential
privacy and $\mathcal{O}(1/\sqrt{T})$ convergence.
Read the full abstract ▾
All authors Quoc H. Nguyen, Ali Lafzi, Abhijeet Phatak, Siddharth Pratap Singh, Rohit Upadhyay, Yogananda Domlur Seetharama, Chittaranjan Tripathy