Artificial Intelligence
Beyond expert users: agents should help users construct preferences, not just elicit them
Irena Saracay, Ludwig Schmidt and Carlos Guestrin
Agents typically assume an expert user – one with well-formed preferences about what they want – and default to clarifying questions whenever the task is underspecified. We argue this assumption is unrealistic. Users often lack the domain knowledge to have completely specified preferences; if asked about their preference on some feature, the user may be unable to answer without the
agent helping the user to learn some domain knowledge needed to form a preference for that feature, e.g., via examples or explanations. To formalize these principles, we draw on the Search-Experience-Credence framework from Information Economics to introduce CoPref, a model of how users construct preferences based on
agent dialog actions. We then study these ideas concretely in agentic recommender systems, proposing CoShop, an interactive
benchmark. In CoShop, an
agent converses with and makes recommendations for a CoPref user. The
agent's performance depends on whether it can help the user gain the knowledge needed to specify the task well. Evaluating five frontier models, we find that no
agent exceeds 56% accuracy on CoShop despite five turns of interaction. Failures stem not from
agents' ability to find items, but from how little the interaction expands what users know about what they want.
Read the full abstract ▾
Artificial Intelligence
When Does Learning to Stop Help? A Cost-Aware Study of Early Exits in Reasoning Models
Zhe Dong, Fang Qin and Manish Shah
Reasoning models spend test-time compute unevenly across instances, and a growing family of early-exit rules – confidence thresholds,
entropy monitors, answer-stability checks, and learned stoppers – promises to reclaim the waste. These rules, however, are evaluated under heterogeneous protocols that leave the deployment question unanswered: at a fixed tolerance for losing correct answers, which policy saves more compute, and does the saving survive probe overhead? We answer this question with a controlled study across 18 task-model settings spanning GSM8K, MATH-500, MMLU-Pro, AIME-90, and GPQA on Qwen3 and DeepSeek-R1-distilled models, using LearnStop, a hidden-state-free logistic stopper over prefix-observable features, as the learned policy instrument. Under matched lost-correct risk at $\alpha$ = 0.15, with the scalar competitor selected on calibration data from confidence,
entropy, confidence-leap, and run-stability exits, the answer forms three regimes. Learned stopping wins on all four primary Qwen3 free-form math settings (+3.2 to +21.2 pp additional total-
token saving); calibrated scalar exits win on multiple-choice MMLU-Pro; and small hard
benchmarks (AIME-90, GPQA) admit no certifiable aggressive policy at all. A trajectory decomposition predicts the regime: learning pays where answers oscillate and correctness evidence is spread across complementary signals, while a single confidence threshold suffices where most instances are already solved at the first checkpoint. Cost accounting sharpens the picture further – the same policy that saves 32% of
tokens under KV-cache forking costs 121% extra under black-box repeated prefilling. Together, these results replace the single-method race with a decision procedure for choosing a stopping rule from the trajectory structure and serving regime of the target workload.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.CL
Test-Time Verification for Text-to-SQL via Outcome Reward Models
Mattia Tritto, Giuseppe Farano, Dario Di Palma et al. (+4 more)
Improving the reliability of large
language models (LLMs) at inference time is a central challenge in structured
reasoning tasks such as Text-to-SQL. Common test-time inference strategies, including Best-of-N sampling and Majority Voting, rely on heuristic signals such as execution success or output frequency, which provide limited semantic discrimination across candidate outputs. In this work, we study Outcome Reward Models (ORMs) as learned semantic scoring functions for test-time verification in Text-to-SQL. While ORMs have been previously explored for test-time scaling and
alignment, their application to structured query generation remains underexplored. We introduce GradeSQL, a scalable framework for training task-specific ORMs via automated candidate generation and execution-based labeling, enabling verifier training without manual annotation. We integrate ORMs into a verification-driven Best-of-N pipeline and evaluate our approach on the BIRD and Spider
benchmarks across multiple open-source LLM families. ORM-based selection consistently outperforms execution-based Best-of-N and Majority Voting, with gains of up to +4.33% on BIRD and +2.10% on Spider. We further show that ORMs scale effectively with larger candidate sets and yield stronger improvements on complex queries. Overall, our results demonstrate that ORM-based verification provides a simple, effective, and scalable alternative to heuristic test-time selection strategies for Text-to-SQL. Code
datasets and models are publicly available.
Read the full abstract ▾
All authors Mattia Tritto, Giuseppe Farano, Dario Di Palma, Gaetano Rossiello, Fedelucio Narducci, Dharmashankar Subramanian, Tommaso Di Noia
Artificial Intelligence
BayesBench: Evaluating LLM Belief Trajectories Under Multi-Turn Evidence Accumulation
Ankur Samanta, Akshayaa Magesh, Tal Lancewicki et al. (+7 more)
Large
language models (LLMs) are typically deployed in multi-turn conversations, where each turn provides new evidence that should reduce epistemic uncertainty about their environment. Acting rationally then requires inferring the unobserved quantities that govern it and updating beliefs about them as evidence accumulates. Yet most evaluations only score the model's final-turn answer in a single-turn format, leaving this process unexamined. We ask how closely LLMs' belief updates match those of a rational
Bayesian reasoner in multi-turn settings, and introduce BayesBench, a suite of simulation environments that probe this across three progressively complex tasks: (i)
Bayesian estimation, where the model infers an unknown parameter from sequential evidence; (ii)
Bayesian prediction, where the model turns inferred beliefs about a latent variable into outcome forecasts; and (iii) latent-framed
Bayesian prediction, where observations are filtered through a user-persona framing, requiring joint inference over the latent state and the persona. Across seven LLMs (3B–70B), scaling improves latent inference and evidence accumulation, with updates occasionally matching the
Bayesian posterior. However, these gains do not reliably carry over to downstream prediction, exposing a gap between inferring latent structure and using it to rationally update beliefs about the target outcome.
Read the full abstract ▾
All authors Ankur Samanta, Akshayaa Magesh, Tal Lancewicki, Ayush Jain, Youliang Yu, Paul Sajda, Kaveh Hassani, Aditya Modi, Daniel R. Jiang, Yonathan Efroni
Artificial Intelligence
How Can AI Find My Model? A Model-Finding Experimental Study Considering Data Formats, Embeddings, and Retrieval Strategies
Jhon G. Botello, Jose J. Padilla, Erika Frydenlund et al. (+2 more)
Discovering simulation models for reuse remains a fundamental challenge in Modeling and Simulation (M&S). When many models coexist, identifying those that align with a given modeling intent remains difficult. Recent advances in Artificial Intelligence (AI), particularly
retrieval-based approaches, offer a promising pathway to operate at this semantic layer. In this paper, we present an experimental study investigating the impact of data representation,
transformer-based
embedding models, and
retrieval strategies on the discovery of simulation models using natural
language queries. We evaluated performance across multiple query types using standard information
retrieval metrics, including recall@5 and nDCG@5. Results show that data representation matters, open-source
embedding models can achieve high performance, and reranking methods are important, especially as query complexity increases. This work provides a baseline for AI-driven model discovery and discusses its role in advancing toward AI-driven composability and interoperability.
Read the full abstract ▾
All authors Jhon G. Botello, Jose J. Padilla, Erika Frydenlund, Krzysztof Rechowicz, Eric Weisel
Artificial Intelligence
Contrastive Reflection for Iterative Prompt Optimization
Derek Koh, Jinghui Mo, Benjamin H. Le et al. (+7 more)
LLM
agents are becoming central to information
retrieval: they issue
retrieval queries, synthesize answers, and increasingly serve as judges for IR evaluation. Improving the
prompts that control these
agents is an optimization problem, but in applied IR settings it often looks less like blind search and more like debugging. Engineers need to know which behavior failed, which nearby behavior still worked, what distinguishes the two, and whether a
prompt edit improves held-out quality without introducing regressions. We present Contrastive Reflection, an iterative
prompt-optimization framework for agentic IR workflows. The framework starts from a task-centric quality definition: QA
agents expose
retrieval or
reasoning traces, and grading
agents expose dimension-level scores and rationales. These structured traces are used to identify error-anchored behavioral slices, add nearby successful examples from the same region, and ask a Teacher LLM to propose a targeted
prompt edit. Candidate edits are accepted only when validation performance improves, optionally subject to regression checks. We instantiate the framework with a tree-based slice selector, but the contribution is the contrastive reflection loop rather than the tree itself. On a public HotpotQA
retrieval-augmented QA setup, one tree-selected contrastive repair improves held-out exact-match accuracy from 51.4% to 60.4%. Failure-only and random-evidence variants improve less and break more previously correct examples. A light instruction-only comparison places the method near modern
prompt optimizers: MIPROv2 reaches 59.4% and GEPA 57.0%. The result is an interpretable optimization loop for IR
agents, aimed at making
prompt repair more inspectable and validation-driven.
Read the full abstract ▾
All authors Derek Koh, Jinghui Mo, Benjamin H. Le, Jiening Zhan, Baofen Zheng, Kevin Bevis, Nathaniel C. Owen, Lauren Elizabeth Charney, Wenqiong Liu, Jingwei Wu
Artificial Intelligence · Cross-listed from cs.LG
A Stationary-Distribution Theory for Triplet-Based Plateau Search in Random Forest Ensemble-Size Selection
Andrey A. Dukhovny and Andrey M. Lange
The number of trees is a central computational parameter in Random Forests: increasing it reduces finite-ensemble variability but increases training and prediction cost. Plateau-based tuning adapts this parameter through local comparisons of out-of-bag scores at a
geometric triplet of tree counts. After the remaining hyperparameters have stabilized, however, the central triplet point need not converge to a deterministic value; instead, it fluctuates around a stationary regime. This paper develops a stationary-distribution theory for this process. The central ensemble size $B_t$ is modeled as a birth-death Markov chain on a
geometric grid, and its stationary distribution is derived through local balance. Under a leading centered folded-normal approximation,
equilibrium equations are obtained for the original update rule and a symmetric modified variant, implying that the stationary center $B_*=O(\varepsilon^{-2})$ as $\varepsilon\downarrow 0$. The stationary spread is also characterized. A local Gaussian approximation and a Fokker-Planck interpretation give grid-level variance constants. After conversion to the ensemble-size scale, $\sigma_{B,*}=O(\varepsilon^{-2})$, while the variance is $O(\varepsilon^{-4})$. The leading relative spread is independent of $\varepsilon$ and controlled by the scale factor and update rule. These results interpret plateau-based Random Forest tuning as a
stochastic process rather than a deterministic stopping rule.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.CR
AI-Generated PowerShell Malware: An Experimental Framework and Dataset
Luciano Pianese, Vittorio Orbinato, Pietro Liguori et al. (+1 more)
Generative AI has emerged as a significant cybersecurity threat, with several recent attack campaigns leveraging LLMs to generate code for malicious purposes via scripting
languages such as PowerShell. Consequently, for cybersecurity analysts, it is imperative to investigate the offensive capabilities of AI code generators. In this paper, we propose an experimental framework to assess LLM-generated PowerShell malware, which comprises a novel sandbox approach for dynamic analysis of AI-generated malware. Furthermore, we present a novel, manually curated
dataset of real-world PowerShell malware, annotated in natural
language to assist the training and evaluation of LLMs. Finally, this study evaluates permissive, open-weight LLMs adapted to PowerShell malware generation. Our results reveal a high degree of similarity between real malware and LLM-generated ones in terms of triggered OS malicious events, with a median Jaccard index of 84.5% and 48.4% of instances achieving complete overlap.
Read the full abstract ▾
All authors Luciano Pianese, Vittorio Orbinato, Pietro Liguori, Roberto Natella
Artificial Intelligence · Cross-listed from cs.CL
When transformers learn "impossible" languages, what do they learn?
Ram Janarthan, Coleman Haley and Sharon Goldwater
Recent work suggests that
transformer language models show a bias towards human
languages over unnatural ("impossible")
languages argued to be unacquirable by humans. However, this literature has largely based these claims on differences in sample efficiency and test-set perplexity, rather than on direct evaluations of the linguistic capacities that could plausibly explain non-attestation in human
languages. We evaluate two theoretically motivated linking hypotheses: impossibility arising from deficiencies in grammatical sensitivity or generative production. Using GPT-2 style models trained on perturbed "impossible" variants of English, we measure sensitivity to grammaticality using BLiMP minimal pairs, finding that model performance exhibits only gradual degradation, mediated by the
language's information locality. In contrast, these models exhibited pronounced failures in generation, producing substantially fewer high-quality sentences at longer lengths. Together, these results suggest generative deficiency and transmission failures as a plausible linking hypothesis between
language model behaviour and non-attestation of impossible
languages.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
Gradient Smoothing: Coupling Layer-wise Updates for Improved Optimization
Haoming Meng, Anton Sugolov and Vardan Papyan
Deep neural networks with repeated architectural blocks, such as
transformers, often exhibit structured relationships across layers that emerge during training. Motivated by this observation, we introduce Depth-wise Gradient Augmentation, a general optimization paradigm in which the update applied to each layer is obtained by transforming the collection of block-wise optimizer updates along the depth dimension. Within this framework, we study Gradient Smoothing, a family of depth-wise smoothing methods, and instantiate it with a simple local Window Smoothing operator. The resulting method operates directly on block-wise updates produced by arbitrary base optimizers (e.g., SGD, Adam,
Muon), incurs minimal computational overhead, and is compatible with existing optimization pipelines. We evaluate Gradient Smoothing across a diverse set of architectures and training regimes, including
language model pretraining, RL post-training of LLMs for
reasoning,
diffusion modeling, and image classification with
Vision Transformers. Across these settings, Gradient Smoothing consistently improves optimization and generalization performance without modifying model architectures or training objectives. We further show that it promotes more structured representation
evolution across depth, consistent with its interpretation as a structured depth-wise preconditioning method. Together, these results establish Depth-wise Gradient Augmentation as a promising framework for exploiting cross-depth structure in optimization and demonstrate Gradient Smoothing as a simple and broadly applicable instantiation.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.CL
Indi-RomCoM: Code-Mixed Benchmark for Evaluating LLMs on Romanized Indic-English Instructions
Avisha Das, Mihir Parmar, Mohana Ramnath et al. (+1 more)
Romanized Code Mixing (RCM), where bilingual speakers fluidly blend local
languages with English in Roman script, has emerged as the dominant form of communication across
multilingual communities. While Large
Language Models (LLMs) perform strongly on monolingual and native-script
benchmarks, their ability to follow instructions and reason over RCM-based content remains largely unexplored. To this end, we introduce the Indi-RomCoM
benchmark for facilitating systematic evaluation on Indic Romanized Code-Mixed instructions. Our
benchmark spans seven instruction-following tasks, four widely spoken Indic
languages, and three controlled code-mixing intensity levels. We extensively evaluate a suite of LLMs covering proprietary, open-weight, and Indic-focused models under zero- and few-shot settings. LLMs consistently underperform on RCM instructions, with performance degrading as code-mixing density increases. Furthermore,
reasoning tasks suffer less degradation than detection tasks (e.g., Toxicity) because the generated explanations offer necessary context. We believe Indi-RomCoM helps the community in developing inclusive
multilingual systems.
Read the full abstract ▾
All authors Avisha Das, Mihir Parmar, Mohana Ramnath, Pulkit Verma
Artificial Intelligence · Cross-listed from cs.CR
Security–Fidelity Tradeoffs: The Hidden Cost of Prompt Injection Defense
Mitchell Hermon, Rahul Gupta, Weitong Ruan et al. (+2 more)
We identify a security-fidelity tradeoff in defending LLMs against indirect
prompt injection: defenses resist injected instructions largely by suppressing untrusted text, which corrupts tasks that must preserve it, such as translation and document editing. Attack-success metrics cannot see this, because a model that ignores an injection and one that faithfully processes it as data score identically. We introduce SecFid, a
benchmark built so that executing an injection, processing it as data, and ignoring it produce distinguishable outputs. This makes fidelity measurable and exposes a frontier: across 1,168 examples and 48 configurations, no model or defense achieves both objectives. The highest-fidelity model reaches 96.5% fidelity at 47.8% security, while the most secure defenses invert this, at 99.3% security but only 71.0%-73.9% fidelity. Even defenses with identical security differ in how they earn it: some repair hijacks into faithful processing, others simply suppress benign content. A decision-theoretic analysis shows why no fixed choice can be right everywhere: the correct behavior is not a property of the defense but of the deployment, set by its relative cost of a hijack versus a dropped span. Security alone therefore measures only half of robustness, and reporting it without fidelity hides the price at which it was bought.
Read the full abstract ▾
All authors Mitchell Hermon, Rahul Gupta, Weitong Ruan, Ekraam Sabir, Haohan Wang
Artificial Intelligence · Cross-listed from eess.AS
Detecting Audio Deepfakes on the Edge:Lightweight SSL-Based Detection in a Browser Plugin
Octavian Pascu, Dan Oneata, Horia Cucu et al. (+1 more)
Audio deepfakes are a growing challenge for the general public, as well as for journalists and fact-checkers. The latter need reliable tools to verify the authenticity of their sources, while at the same time keeping their information private. Commercial deepfake detection solutions rely on cloud-based processing, which raises
privacy concerns. To solve this problem, we propose an on-device audio deepfake detection model. We show that a truncated self-supervised backbone with a simple logistic classifier is both very fast and often more accurate than existing solutions. Our solution outperforms the baseline AASIST by 10% and improves inference speed by 40%. We integrate this model into a browser plug-in, which allows journalists and fact-checkers to detect deepfakes easily and securely. Code for the plugin is available at https://github.com/OctavianPascu97/Audio-Deepfakes-Browser-Plugin.
Read the full abstract ▾
All authors Octavian Pascu, Dan Oneata, Horia Cucu, Nicolas M. Muller
Artificial Intelligence · Cross-listed from cs.CL
A Single Rewrite Suffices: Empirical Lessons from Production Skill Description Optimization
Yangqiaoyu Zhou, Mohammad Alqudah, Kwei-Herng Lai et al. (+3 more)
Enterprise AI
agents route user queries to specialized skills by matching queries against natural
language skill descriptions. When two skills share overlapping descriptions, the routing LLM misroutes queries, a failure we term skill collision. As
agents scale to dozens of skills, manually tuning descriptions to maintain routing accuracy becomes a significant engineering bottleneck. We deploy an automated description optimization pipeline on a production enterprise group chat
agent (9 skills, 372 regression cases). The pipeline produces descriptions averaging 79.2% F1, matching manually tuned descriptions at 79.4% F1 (average per-skill difference -0.20%, within the 0.78% multi-seed noise floor), while reducing per-skill engineering effort from 120 minutes to 3.8 minutes (32 times speedup). We then examine which pipeline components actually drive this match. Systematic ablation on both the production system and ToolBench (16k tools) reveals that a single LLM rewrite using any available false-positive and false-negative cases captures most of the available improvement. Other design choices we tested (iteration budget, feedback signal composition, dual editing of confused pairs, and training set size) each affect final F1 by less than 0.5%. Description optimization addresses skill collisions caused by overlapping descriptions but cannot resolve cases where two skills intended scopes genuinely overlap. We identify a diagnostic (a large train-validation F1 gap) that flags the latter cases for architectural rather than text-level intervention.
Read the full abstract ▾
All authors Yangqiaoyu Zhou, Mohammad Alqudah, Kwei-Herng Lai, Aaron Halfaker, Yingqi Xiong, Yaar Harari
Artificial Intelligence
What Drives Interactive Improvement from Feedback?
Bartłomiej Cupiał, Jan Łojek, Mikołaj Garstecki et al. (+3 more)
We study when natural-
language feedback produces improvement beyond the gains obtainable from repeated attempts alone. In multi-turn
language agent setting, higher final accuracy can reflect useful feedback, but it can also arise from resampling, format correction, or additional test-time computation. To separate these effects, we introduce a controlled student-teacher protocol across Omni-MATH, Codeforces, BBEH Linguini, and ARC-AGI1, evaluating thirteen open-weight models in both student and teacher roles. We compare external feedback, self-feedback, and unguided self-refinement, while varying interaction history, task difficulty, and teacher access to privileged task information. Across settings, we find that multi-turn improvement is often not evidence of feedback use: self-generated feedback adds little beyond unguided self-refinement, whereas the strongest external teachers produce substantially larger feedback-specific gains, suggesting that useful feedback must provide guidance beyond generic retry. Dense student-teacher interaction
matrices further show that interactive gains are driven more by the student's ability to use feedback than by the teacher's identity, although teacher choice remains important for a fixed student. These results suggest that feedback-based
agents should be evaluated against repeated-attempt baselines, and that ability to act on feedback, not merely feedback availability, is a central bottleneck for interactive improvement. We release our controlled student-teacher evaluation framework at https://j-lojek.github.io/feedback-generation-is-a-bottleneck/.
Read the full abstract ▾
All authors Bartłomiej Cupiał, Jan Łojek, Mikołaj Garstecki, Szymon Pobłocki, Alicja Ziarko, Piotr Miłoś
Artificial Intelligence · Cross-listed from cs.CR
Understanding and Evaluating Claw-like Agent Security Through a Computer-Systems Lens
Peizhi Niu, Wenjie Qu, Shangding Gu et al. (+14 more)
Claw-like AI
agents (e.g., OpenClaw) are always-on processes with persistent access to credentials, files, tools, and external services. They take on system-level responsibilities – installing packages, maintaining state, scheduling subtasks, and mediating I/O – making security failures far more severe than in other
agents. Yet existing
benchmarks focus on model responses and tool calls, leaving cross-component failure modes largely unmeasured. We adopt a computer-system analogy: treating a Claw-like
agent as an agentic computer system whose gateway runtime plays an OS-like mediation role, whose Skills resemble user-installed applications, and whose Plugins resemble loadable extensions with runtime privileges. Each component has a classical counterpart whose protection mechanisms – refined over decades of cybersecurity research – are absent on the
agent side. From this perspective, we develop SafeClawArena, a
benchmark of 406
adversarial tasks across four attack surfaces (Skill Supply-Chain Integrity, Persistent State Exploitation, Cross-Boundary Data Flow, and Indirect
Prompt Injection), executed in containerized replicas of real
agent platforms with canary-marked credentials and evaluated via automated taint tracking across nine output channels. We evaluate three platforms (OpenClaw, NemoClaw, SeClaw) and five frontier LLMs. The highest attack success rate reaches 70%; malicious Plugins succeed in 100% of cases regardless of the LLM. SeClaw cuts GPT-5.4's attack success rate from 70% to 22%, partly through utility-security tradeoffs rather than active defenses, while Claude-Opus-4.6 already sits near a 22% floor on every platform. These results expose the inadequacy of current defenses and suggest directions for future hardening. Code and data: https://github.com/sunblaze-ucb/SafeClawArena.
Read the full abstract ▾
All authors Peizhi Niu, Wenjie Qu, Shangding Gu, Tianneng Shi, Yuankai Li, Ahmad Tawaha, Hend Alzahrani, Vincent Siu, Boyi Li, Chenguang Wang, Jiaheng Zhang, Basel Alomair, Ming Jin, Muhao Chen, Chi Wang, Costas Spanos, Dawn Song
Artificial Intelligence · Cross-listed from cs.LG
Hierarchical Global Attention (HGA)
Woernle Frank, Fedosov Vladimir and Grinenko Artemiy
Hierarchical Global
Attention (HGA) is a drop-in replacement for dense causal
attention in pretrained long-context
transformers. HGA preserves the original checkpoint parameters: the pretrained $W_Q$, $W_K$, $W_V$, and $W_O$ projections remain unchanged, no calibration parameters are introduced, and no retraining is required. Applied to Qwen3-30B-A3B-Instruct-2507-FP8 on a single RTX~5090 (32GB), the patched model runs out of the box at a 64K-
token context, where
token-level K/V storage is not feasible on this hardware. Unlike previous sparse-
attention methods, HGA performs hierarchical two-level routing. It first retrieves relevant chunks using compact RoPE-aware summaries and then refines the selection by routing only the most relevant groups before performing exact
token-level
attention. This hierarchical
retrieval significantly reduces the number of fetched
tokens while preserving exact
attention over the retrieved
token set, making RAM- and NVMe-backed storage practical. The full historical
token K/V resides in host RAM or NVMe storage, while only a small routed working set is transferred to GPU
memory during
attention. Consequently, GPU
memory consumption depends primarily on model weights and the routed working set rather than on the total context length. Across all tested context lengths (4K - 64K
tokens), routed
attention remains within approximately $0.01$–$0.02$ nats of dense
attention while the sparsity used is just about 3%. These results suggest that the approximation introduced by hierarchical routing is small, and that the remaining quality gap is likely dominated by long-context positional encoding rather than by the routing
algorithm itself.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
Why Do Few-Step Text Latents Fail When Image Latents Work? Non-Commitment at Sharp Categorical Readouts
Zhongyao Wang
Deterministic few-step generation succeeds on continuous image latents but collapses to incoherent text on continuous text latents, and we show the cause is
geometric rather than a training or scaling deficiency: a smooth, regularity-limited deterministic map cannot resolve a discrete branch choice before a sharp categorical readout, so few-step failure is governed by decoder sharpness, not transport accuracy. In the overlapping regime of real text autoencoders, we prove (
Theorem 3) that the posterior-mean terminal step flips
tokens at the rate of the latent mass in an $O(s(t))$ tube around decision boundaries. Two diagnostics, DABI (readout sharpness) and CCI (categorical commitment), measured on published checkpoints show that four independently built continuous-text decoders amplify a boundary-aligned perturbation far beyond a norm-matched isotropic one (DABI from $5\times10^{2}$ to $>10^{5}$), while image decoders have DABI $\approx 1$. Two mechanisms escape the continuous bound: categorical commitment (autoregressive decoders succeed despite sharper readouts) and
stochastic re-injection (deterministic ODE at $K=4$ gives PPL 294 versus SDE 50 on the same model). In the idealized separated regime we prove matching sharp transport laws, including a dimension phase diagram: the deterministic stiffness needed to separate $M$ modes grows as $\Theta(\sqrt{\log M})$ once the latent dimension is $\Omega(\log M)$ (and as $M^{1/n}$ in fixed dimension), with a depth-$B$ hierarchy giving a $\sqrt{B}$-smaller per-step peak (
Theorems 5-7); a coarea identity links these to the overlapping tube (
Theorem 17). The result is an accuracy-depth-stiffness tradeoff: within the deterministic-continuous class the cost is irreducible, and both escapes step outside it.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
From Search to Synthesis: Training LLMs as Zero-Shot Workflow Generators
Gan Luo, Zihan Qin, Bin Dong et al. (+1 more)
Large
language models (LLMs) excel across a wide range of tasks, yet their instance-specific solutions often lack the structural consistency needed for reliable deployment. Workflows that encode recurring algorithmic patterns at the task level provide a principled framework, offering robustness across instance variations, interpretable traces for debugging, and reusability across problem instances. However, manually designing such workflows requires significant expertise and effort, limiting their broader application. While automatic workflow generation could address this bottleneck, existing methods either produce instance-specific solutions without learning task-level patterns, or cannot generalize beyond their training configurations. We present MetaFlow, which casts workflow generation as a meta-learning problem: given a task and an operator set, the model learns to compose solution strategies. MetaFlow trains in two stages: supervised fine-tuning on synthetic workflow data, followed by reinforcement learning with verifiable rewards (RLVR) that uses execution feedback across problem instances in the task to improve end-to-end success. The resulting model produces effective workflows for trained tasks and exhibits strong generalization to untrained tasks and novel operator sets. Across
benchmarks in question answering, code generation, and mathematical
reasoning, MetaFlow achieves performance comparable to state-of-the-art baselines on in-domain tasks with single inference, while demonstrating remarkable zero-shot generalization capabilities on out-of-domain tasks and operator sets.
Read the full abstract ▾
All authors Gan Luo, Zihan Qin, Bin Dong, Wotao Yin
Artificial Intelligence · Cross-listed from cs.LG
Accelerometry-Derived Digital Biomarkers for Cardiometabolic Risk: A Population-Representative Tabular Benchmark with Uncertainty Quantification
Federico Felizzi
Structured tabular data dominates clinical medicine, yet existing
benchmarks fail to reflect real-world properties like complex survey sampling, demographic oversampling, and subgroup
fairness. We introduce the NHANES Accelerometry Cardiometabolic
Benchmark, derived from NHANES 2003-2006, comprising 1,381 adults with hip-worn accelerometry, fasting laboratory biomarkers, dietary intake, and anthropometrics. We evaluate three tabular learning methods – ridge regression, XGBoost, and the foundation model TabPFN v2 – to predict glycated haemoglobin (HbA1c), fasting triglycerides, and C-reactive
protein (CRP) from activity phenotypes and lifestyle
covariates. TabPFN v2 achieves the best overall performance (HbA1c R^2=0.156, CRP R^2=0.383), while triglycerides remain largely unpredictable (R^2 < 0.05), consistent with known genetic dominance. We apply split conformal prediction to generate distribution-free 90% prediction intervals and evaluate demographic coverage equity across sex and race/ethnicity subgroups. Marginal coverage aligns with the 90% target for CRP and HbA1c but falls below for triglycerides. At the subgroup level, we observe localized undercoverage (e.g., HbA1c for Mexican American participants), illustrating the gap between marginal guarantees and the conditional coverage required for clinical
fairness. Code and data are at https://github.com/felizzi/nhanes-accel-cardiometabolic-
benchmark.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.CR
An AI-Based Solution for Secure Service Provisioning in IoT
Marco Arazzi, Mert Cihangiroglu, Serena Nicolazzo et al. (+2 more)
As the Internet of Things (IoT) continues its rapid expansion, the attack surface grows accordingly, with emerging threats targeting smart objects and their interactions. In this evolving landscape, securing service provisioning is crucial to ensure the proper functioning, security, and reliability of the IoT ecosystem. Service provisioning encompasses key tasks such as device registration, configuration, authentication, authorization, and software deployment, all of which are essential for seamless and secure IoT operations. In this paper, we present a comprehensive framework designed to select the most suitable smart objects to deliver a target service within a given IoT environment while also monitoring the behavior of the entities involved during the service provisioning phase. To achieve this, we employ a Deep Reinforcement Learning (DRL) approach in which an intelligent
agent learns, through interaction with a complex, dynamic environment, how to adapt to changes while adhering to predefined security constraints. For behavioral monitoring, we leverage
Federated Learning (FL) to develop a global Behavioral Fingerprinting (BF) model that is fully distributed and can analyze how IoT devices interact within the network. In addition, the BF is used to compute a reliability score for each service provider, reflecting its degree of compliance with the defined security constraints. This score is then incorporated into the service provisioning process, allowing smart objects to select providers not only according to functional suitability but also to their reliability level. Finally, we conduct an extensive experimental evaluation to assess the robustness and scalability of our approach. The results demonstrate that our solution can be effectively deployed even on resource-constrained IoT devices, making it a viable and scalable security-enhancing mechanism for modern IoT ecosystems.
Read the full abstract ▾
All authors Marco Arazzi, Mert Cihangiroglu, Serena Nicolazzo, Antonino Nocera, Vinod P
Artificial Intelligence · Cross-listed from cs.SD
BEST-RQ-2: Contextualize-Then-Predict, a Two-Step Approach for Self-Supervised Audio Representations
Ludovic K. Tuncay, Etienne Labbé and Thomas Pellegrini
Self-supervised learning enables audio representations that transfer across domains and tasks. We present BEST-RQ-2, an
evolution of BEST-RQ that retains frozen randomprojection-based discrete targets while introducing a two-step contextualize-then-predict pretraining scheme. A ViT context encoder processes only the unmasked spectrogram regions, and a lightweight predictor infers targets for the masked regions; the predictor is discarded after pretraining. Replacing the original Conformer encoder with a ViT shifts performance across domains, slightly reducing
speech performance while improving music and environmental sounds, with comparable average scores. The main improvement comes from decomposing masked prediction into separate contextualization and prediction stages. On the X-ARES and XARES-LLM
benchmarks, BEST-RQ-2 consistently outperforms one-stage baselines in overall transfer while keeping inference compute unchanged. Code and model checkpoints are publicly available.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.OS
LUMOS: A Semantic Operating-System Layer for Accessibility-Grounded AI Agents
Yogeswar Reddy Thota
Current operating systems expose interfaces optimized for human users but not for AI
agents. Humans benefit from pixels, icons, windows, visual grouping, mouse movement, and keyboard shortcuts; AI
agents instead need compact semantic state, grounded actions, and reliable feedback. As a result, many computer-use
agents are forced to interpret screenshots, OCR output, and visual crops, introducing high
token costs, visual ambiguity, latency, and coordinate uncertainty. This paper introduces LUMOS (
Language Model Unified Machine-Readable Operating-System Semantics), a semantic interaction layer between AI
agents and operating systems. LUMOS converts native accessibility metadata and browser UI structures into machine readable semantic blueprints with stable identifiers, roles, names, values, bounds, and action affordances. It also supports live semantic pointer grounding by querying the UI element under or near the cursor through operating-system automation APIs. An LLM then acts through an accessibility grounded observe act loop using constrained visible-UI primitives rather than application-specific scripts. LUMOS does not claim to replace visual
agents; instead, it reduces dependence on screenshots when operating systems already provide semantic structure. These results suggest a path toward AI-native operating systems and machine-readable interaction layers.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from q-bio.QM
Modeling Cell-Cycle-Aware Single-Cell Drug Perturbation Responses
Dingping Zhao, Jie Lin, Feng Xu et al. (+1 more)
Single-
cell drug perturbation models should capture transcriptional response magnitude and whether a treatment changes the proliferative state of the
cell. This is difficult because
cell-cycle variation is often treated as a nuisance factor, and
benchmark processing rarely makes drug-induced phase changes a primary prediction target. We introduce scCycleMol, a
cell-cycle-aware perturbation prediction framework built on a curated 24-hour SciPlex3
benchmark with standardized molecule identities, dose and
cell-line metadata, modeled genes, and expression-derived
cell-cycle supervision. scCycleMol derives
cell-cycle supervision from the treated state and applies it to predicted treated expression without using phase as an input
covariate. The model includes a learnable full-expression
cell-cycle head with circular G1/S/G2M targets, and we evaluate readout-only supervision (with stop-gradient) versus closed-loop supervision (backpropagating through decoder, dose-response module, and drug representation). We also compare molecular representations and pretraining sources to isolate the effect of the
cell-cycle objective. On a processed 24-hour SciPlex3
benchmark (635,541
cells, 186 perturbations, 188 compound
embeddings, 3
cell lines, 4 doses plus DMSO, 5,080 genes), the best LINCS-pretrained circular variant reaches 0.9093 mean all-gene R-squared and 0.6843 mean DE-gene R-squared. Under matched preprocessing, closed-loop
cell-cycle supervision improves phase accuracy by 0.54-0.62 points while keeping mean all-gene R-squared within 0.003 of matched chemCPA no-
cell-cycle models; Tahoe-pretrained readout-only circular supervision achieves the strongest phase accuracy at 0.9609.
Read the full abstract ▾
All authors Dingping Zhao, Jie Lin, Feng Xu, Zhengwei Xie
Artificial Intelligence · Cross-listed from cs.RO
DSIP: A Dynamic Coordination Planner for Signal-Free Intersections using Diffusion-Model-Based Multi-Agent Motion Planning
Qian Hu, Haoyang Peng, Songan Zhang et al. (+2 more)
Traffic signal control at urban intersections inherently introduces stop-and-go behavior, resulting in increased delays and reduced traffic efficiency, especially under high traffic demand. With the emergence of connected and automated vehicles (CAVs), trajectory-level coordination has emerged as a high-potential strategy to augment or transcend conventional phase-based management. This paper proposes DSIP (
Diffusion-model-based Signal-free Intersection Planner), a multi-
agent motion planning framework driven by a generative
diffusion process. DSIP shifts the intersection management paradigm from discrete temporal phasing to continuous multi-vehicle trajectory optimization. This work evaluates the theoretical upper-bound performance of this coordination strategy under idealized communication and execution conditions to isolate the core benefits of the
diffusion-driven approach. Using the SUMO platform, we evaluate DSIP across diverse four-leg intersection configurations. Experimental results demonstrate that DSIP significantly reduces average delay and maintains higher average speed compared to both fixed-time signal control and state-of-the-art reinforcement-learning-based controllers, particularly in medium- to high-density traffic. These findings suggest that
diffusion-based trajectory planning provides a scalable and robust foundation for future autonomous intersection management. By unlocking latent intersection capacity through software-defined coordination, this approach offers a cost-effective pathway to improve urban traffic flow efficiency without requiring physical infrastructure expansion.
Read the full abstract ▾
All authors Qian Hu, Haoyang Peng, Songan Zhang, Ming Yang, Hongtei Eric Tseng
Artificial Intelligence · Cross-listed from cs.RO
VLK: Learning Humanoid Loco-Manipulation from Synthetic Interactions in Reconstructed Scenes
Yen-Jen Wang, Jiaman Li, Sirui Chen et al. (+9 more)
Perception-based
humanoid loco-manipulation requires connecting
egocentric observations and task instructions to whole-body motion. Learning this mapping requires synchronized
egocentric images,
language commands, and
robot-compatible kinematic trajectories, yet no existing data source provides this complete tuple at scale. We address this bottleneck by generating
vision-
language-kinematics (VLK) supervision synthetically in reconstructed scenes. Our pipeline leverages 3D Gaussian Splatting to reconstruct metric-scale indoor environments, synthesizes
navigation and object-interaction trajectories using privileged scene information, and renders paired
egocentric observations after the fact. We produce 48,000 paired trajectories with no human intervention and train a VLK policy that predicts short-horizon whole-body kinematic trajectories. A whole-body tracker converts these predictions into actions on the physical
humanoid. We evaluate on the physical Unitree G1 performing
navigation and single-object transport, demonstrating that synthesized interactions in reconstructed scenes provide effective supervision for sim-to-real perception-based
humanoid loco-manipulation. Project Website: https://
vision-
language-kinematics.github.io/
Read the full abstract ▾
All authors Yen-Jen Wang, Jiaman Li, Sirui Chen, Takara E. Truong, Pei Xu, Pieter Abbeel, Rocky Duan, Koushil Sreenath, Angjoo Kanazawa, Carmelo Sferrazza, Guanya Shi, Karen Liu
Artificial Intelligence · Cross-listed from cs.SD
LeVo 2: Stable and Melodious Song Generation via Hierarchical Representation Modeling and Progressive Post-Training
Shun Lei, Huaicheng Zhang, Dapeng Wu et al. (+8 more)
Full-length song generation must preserve coherence and musicality, render detailed vocal and accompaniment acoustics, and follow lyrics and
prompts. Existing
language model-based systems face a structural trade-off: mixed-
token modeling preserves vocal-instrument coordination but obscures track-specific details, whereas dual-track prediction improves acoustics but requires longer sequences and weakens global planning. We present LeVo 2, a hybrid LLM-
Diffusion framework for controllable full-length song generation. LeVo 2 formulates this trade-off as hierarchical modeling: LeLM first predicts mixed
tokens for semantic planning, then predicts vocal and accompaniment
tokens in parallel for track-specific refinement, while a
diffusion-based Music Codec reconstructs full-length waveforms. A central contribution of this extended version is an aesthetics-guided training schedule for
alignment. During pre-training, an automated music aesthetic evaluation framework assigns musicality-tier conditions to large-scale data, providing musicality priors before preference
alignment. Progressive post-training applies SFT, large-scale offline DPO, and closed-loop semi-online DPO to separately improve generation quality, controllability, and musicality. Modular extension then trains the Track-Specific LM for acoustic refinement while preserving the aligned semantic planner. This schedule separates musicality learning, controllability
alignment, and acoustic refinement, mitigating optimization conflict and the limitations of static offline preference pairs. Expert listening tests and objective evaluations show that LeVo 2 outperforms open-source baselines across six subjective dimensions, and approaches leading commercial systems on several listening metrics. Ablations validate the effects of the training strategy, aesthetics guidance, scaling, and hierarchical architecture.
Read the full abstract ▾
All authors Shun Lei, Huaicheng Zhang, Dapeng Wu, Yaoxun Xu, Lishi Zuo, Wei Tan, Hangting Chen, Guangzheng Li, Jianwei Yu, Zhiyong Wu, Dong Yu
Artificial Intelligence
Self-Evolving World Models for LLM Agent Planning
Xuan Zhang, Wenxuan Zhang, See-Kiong Ng et al. (+1 more)
World models offer a principled way to equip long-horizon LLM
agents with foresight: predictions of action consequences before execution. However, unreliable foresight can be ignored, misused, or even degrade downstream decision-making. In this paper, we introduce WorldEvolver, a self-evolving world model framework that revises its deployment-time context while keeping the downstream
agent and all model parameters frozen. WorldEvolver integrates three modules: (i) Episodic
Memory, which exploits real action transitions through
retrieval-based simulation; (ii) Semantic
Memory, which extracts persistent heuristic rules from prediction-observation mismatches; and (iii) Selective Foresight, which filters low-confidence predictions before integrating them into
agent reasoning context. We evaluate WorldEvolver on ALFWorld and ScienceWorld, measuring world model prediction accuracy on Word2World and downstream
agent success rate on AgentBoard. Extensive experiments show that WorldEvolver achieves the highest prediction accuracy across three backbones and leads other world model baselines on downstream
agent success rate, demonstrating that test-time
memory revision enhances both predictive fidelity and planning performance.
Read the full abstract ▾
All authors Xuan Zhang, Wenxuan Zhang, See-Kiong Ng, Yang Deng
Artificial Intelligence · Cross-listed from cs.RO
GROW$^2$: Grounding Which and Where for Robot Tool Use
Yuhong Deng, Yuyao Liu and David Hsu
Can the
robot use a plate to cut a cake if no knife is available? Tool use greatly expands
robot capabilities, but to use tools creatively beyond their intended functions, the
robot faces the challenge of $\textit{open-world affordance grounding}$: select an open-category object to act as a tool and localize its specific region of action. To this end, we introduce GROW$^2$ (GROunding Which and Where), which leverages object parts as a natural abstraction to split the grounding process hierarchically into semantic and
geometric levels, thus bypassing the need for data-heavy, end-to-end training. Semantically, GROW$^2$ harnesses the commonsense
reasoning of
Vision-
Language Models (VLMs) to parse a natural-
language task instruction, select a suitable object as the tool, and identify task-relevant parts on the tool and the target object. Geometrically,
vision foundation models then ground the selected parts into precise 3D regions from a single RGB-D image. Experiments on established
benchmarks show that GROW$^2$ outperforms state-of-the-art baselines on affordance prediction
benchmarks. Further, it achieves zero-shot generalization over open-category objects and outperforms baselines in both simulated and real-world
robot tool use experiments.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from cs.LG
Pessimism's Paradox: Conservative Offline Training Amplifies Reward Hacking During Online Adaptation in Reasoning Models
Subramanyam Sahoo, Aman Chadha, Vinija Jain et al. (+1 more)
Conservative offline training is widely advocated as a safe foundation for subsequent online adaptation: if a policy stays close to well-supported behaviour, the argument goes, it is less likely to exploit imperfections in a learned reward model. We challenge this intuition empirically and mechanistically. We train a Qwen3-14B policy under Direct Preference Optimisation (DPO) with three levels of conservatism ($\beta \in \{\beta_{\mathrm{lo}}, \beta_{\mathrm{mid}}, \beta_{\mathrm{hi}}\}$ derived from empirical log-ratio percentiles), then adapt each checkpoint online against a learned reward ensemble (3\,$\times$\,Qwen3-1.7B) while measuring true performance on GSM8K exact-answer accuracy. We find that higher offline conservatism monotonically increases reward-hacking damage, measured by the Goodhart gap and its area under the curve (AUGC), with Spearman $\rho = 1.0$ across all three conditions. Mechanistic analysis reveals a three-link causal chain: (i) high-$\beta$ DPO compresses policy
entropy, (ii) Low-
entropy policies generate responses with reduced diversity, concentrating in a narrow region of the reward model's training distribution (lower pairwise cosine distance), and (iii) despite this proximity, ensemble disagreement (epistemic uncertainty) increases with $\beta$ and is exploited faster during online optimisation. We further fit a power-law curve to the $(\beta, \augc)$ data and identify a practical optimal conservatism level $\beta^{\star}$ that balances
alignment fidelity against hacking vulnerability. Our results suggest that the field needs calibrated, not maximal, conservatism.
Read the full abstract ▾
All authors Subramanyam Sahoo, Aman Chadha, Vinija Jain, Divya Chaudhary