# Research Papers — 2026-05 · bundle page 242 of 326

> Full-text markdown of papers #12051–12100 of 16266 published in 2026-05, newest first. AI-readable mirrors of the full 16266-paper 2026-05 archive (50 papers per file).

Canonical page: https://aipod.dev/papers/archive/2026-05/page/242
[← Previous bundle](https://aipod.dev/papers/archive/2026-05/page/241.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-05/page/243.md)

### MultiMedVision: Multi-Modal Medical Vision Framework

Authors: Frank Li, Bardia Khosravi, Mohammadreza Chavoshi, Young Seok Jeon, Theo Dapamede, Hari Trivedi, Janice Newsome, Judy Gichoya
Published: 2026-05-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2605.09151v1

Multi-modal medical imaging enables comprehensive diagnostics, yet current foundation models process 2D (e.g. X-ray) and 3D (e.g. CT) data with separate, dimensionality-specific architectures. We present MultiMedVision, a unified framework for joint 2D/3D representation learning built on a Sparse Vision Transformer. Our model uses 3D Rotary Positional Embeddings and variable-length sequence packing to process mixed-modality batches natively within a shared latent space, without modality-specific adapters or treating 3D volumes as 2D slice sequences. Trained with a self-supervised objective on chest X-rays (MIMIC-CXR) and CT scans (CT-RATE), and using a single shared encoder with 5x less data, MultiMedVision achieves competitive performance on both 2D benchmarks (Macro AUROC 0.82 on MIMIC, 0.84 on CheXpert) and 3D tasks (0.85 on CT-RATE). Analysis of the learned representations reveals coexisting modality-specific and shared feature subspaces, demonstrating that unified cross-dimensional representation learning is feasible without sacrificing modality-specific performance.

---

### AlphaExploitem: Going Beyond the Nash Equilibrium in Poker by Learning to Exploit Suboptimal Play

Authors: Vlad Murgoci, Matthijs Spaan, Yaniv Oren
Published: 2026-05-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2605.09150v1

Poker is an imperfect information game that has served as a long-standing benchmark for decision-making under uncertainty. To maximize utility beyond the Nash equilibrium, an agent can deviate from Nash-equilibrium policies to exploit suboptimal play. We introduce AlphaExploitem, which extends the competitive RL poker agent AlphaHoldem by using a hierarchical transformer encoder that enables reasoning over previously played hands and modifying the training procedure with the inclusion of a diverse pool of exploitable opponents to facilitate learning to exploit. We train and evaluate AlphaExploitem on two standard benchmarks for imperfect-information games. Empirically, AlphaExploitem successfully exploits weak play by both in- and out-of-distribution opponents, without losing performance against NE opponents.

---

### From Traditional Taggers to LLMs: A Comparative Study of POS Tagging for Medieval Romance Languages

Authors: Matthias Schöffel, Esteban Garces Arias
Published: 2026-05-09
Categories: cs.CL, cs.AI, stat.AP
arXiv: https://arxiv.org/abs/2605.09147v1

Part-of-speech (POS) tagging for Medieval Romance languages remains challenging due to orthographic variation, morphological complexity, and limited annotated resources. This paper presents a systematic empirical evaluation of large language models (LLMs) for POS tagging across three medieval varieties: Medieval Occitan, Medieval Catalan, and Medieval French. We compare traditional rule-based and statistical taggers with modern open-source LLMs under zero-shot prompting, few-shot prompting, monolingual fine-tuning, and cross-lingual transfer learning settings. Experiments on historically grounded datasets show that LLM-based approaches consistently outperform traditional taggers, with fine-tuning and multilingual training yielding the largest improvements. In particular, cross-lingual transfer learning substantially benefits under-resourced varieties, while targeted bilingual training can outperform broader multilingual configurations for specific target languages. The results highlight the importance of linguistic proximity and dataset characteristics when designing transfer strategies for historical NLP. These findings provide empirical insights into the applicability of modern neural methods to medieval text processing and provide practical guidance for deploying LLM-based POS tagging pipelines in digital humanities research. All code, models, and processed datasets are released for reproducibility.

---

### Beyond Thinking: Imagining in 360$^\circ$ for Humanoid Visual Search

Authors: Jingdong Zhang, Yizhou Wang, Zhengzhong Tu, Xin Li, Wenping Wang, Xiaohang Zhan
Published: 2026-05-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2605.09146v1

Humanoid Visual Search (HVS) requires agents to actively explore immersive 360$^\circ$ environments. While prior methods treat this as a monolithic task relying on cumulative, multi-turn Chain-of-Thought (CoT) reasoning, they impose heavy cognitive burdens and require expensive trajectory-level annotations. In this paper, we propose Imagining in 360$^\circ$, a novel framework that decouples the exploration process into a specialized Imaginator and an Actor. The Imaginator functions as a probabilistic predictor of spatial priors; instead of maintaining a cumulative reasoning chain, it infers the semantic layout of both observed and unobserved regions in a single step. By sampling multiple hypotheses within this semantic space, we provide the Actor with a distribution of effective spatial information, offering robust guidance that hedges against uncertainty during active search. This decoupled architecture significantly lowers data engineering costs by eliminating the need for full-trajectory CoT annotations, enabling the generation of over 1.96 million curated training samples. Extensive experiments demonstrate that explicitly modeling semantic spatial priors drastically improves search efficiency and success rates in complex, in-the-wild environments.

---

### FedVSSAM: Mitigating Flatness Incompatibility in Sharpness-Aware Federated Learning

Authors: Bingnan Xiao, Yuan Gao, Bingcong Li, Wei Ni, Xin Wang, Tony Q. S. Quek
Published: 2026-05-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2605.09144v1

Sharpness-aware minimization (SAM) is an effective method for improving the generalization of federated learning (FL) by steering local training toward flat minima. Under data heterogeneity, however, device-side SAM searches for locally flat basins that are incompatible with the flat region preferred by the global objective. We identify this structural failure mode as flatness incompatibility, which explains why improving local flatness alone may provide limited training and generalization improvement for the global model. We reveal that flatness incompatibility arises from data heterogeneity and the friendly adversary phenomenon, and is further amplified by local updates and partial device participation. To mitigate this issue, we propose Federated Learning with variance-suppressed sharpness-aware minimization (FedVSSAM), which constructs a variance-suppressed adjusted direction and uses it consistently in local flatness search, local descent, and global update. FedVSSAM anchors both perturbation and update directions to a more stable global direction, instead of correcting only an isolated local perturbation. We establish non-convex convergence guarantees of FedVSSAM and prove that the mean-square deviation between the adjusted direction and the global gradient is effectively controlled. Experiments demonstrate that FedVSSAM mitigates flatness incompatibility and outperforms the baselines across diverse FL settings.

---

### Evaluating Federated Learning approaches for mammography under breast density heterogeneity

Authors: Gonzalo Iñaki Quintana, Franco Martin Di Maria, Laurence Vancamberg
Published: 2026-05-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2605.09137v1

Breast density is a key factor that influences mammography interpretation and is a major source of heterogeneity in multicenter datasets. Such heterogeneity poses challenges for collaborative machine learning across institutions, particularly in Federated Learning. This study aims to evaluate the impact of breast density-induced heterogeneity on FL for mammography image classification and to assess the robustness of common FL algorithms in realistic clinical settings. We conducted experiments under two scenarios: (1) a strongly heterogeneous setting where each participating site contributed exclusively low- or high-density cases, based on the BI-RADS density score, and (2) a population-based setting simulating breast density distributions in White and Asian populations. For the strongly heterogeneous setting, we evaluated two configurations: one with 2 clients, where the cases were grouped as BI-RADS A-B and C-D, and one with 4 clients, where each site contained cases of a single BI-RADS density. We compared three FL methods (FedAvg, FedProx, SCAFFOLD) against centralized training, local-only training, and naive aggregation approaches, including ensembling and weight averaging. Across both scenarios, FL achieved performance comparable to centralized training, while local models and naive aggregation approaches underperformed in the presence of strong heterogeneity. Notably, FedAvg achieved accuracy on par with or exceeding centralized training, demonstrating resilience to breast density-induced data imbalance without requiring specialized heterogeneity mitigation algorithms. These findings show that FL can address breast density-related heterogeneity, supporting its feasibility for real-world mammography workflows. The demonstrated robustness of FedAvg underscores the potential for broad clinical deployment of FL, enabling collaborative model development while maintaining data privacy.

---

### BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models

Authors: Yuanhao Li, Hongbo Wang, Xiaotang Shang, Xunzhu Tang, Yiming Cao, Xuhong Chen
Published: 2026-05-09
Categories: cs.AI, cs.SE
arXiv: https://arxiv.org/abs/2605.09134v4

Reinforcement learning for program repair is hindered by sparse execution feedback and coarse sequence-level rewards that obscure which edits actually fix bugs. We present BoostAPR, a three-stage framework addressing these challenges: (1) supervised fine-tuning on execution-verified demonstrations with reasoning traces, (2) training dual reward models--a sequence-level assessor and a line-level credit allocator--from execution outcomes, and (3) PPO optimization where the line-level model redistributes rewards to critical edit regions. This line-level credit assignment operates at an intermediate granularity naturally suited to code changes. Trained on SWE-Gym and evaluated on four benchmarks, BoostAPR achieves 40.7% on SWE-bench Verified (+22.9pp over base model), 24.8% on Defects4J (Python-to-Java transfer), 84.5% on HumanEval-Java, and 95.0% on QuixBugs, achieving competitive results among open-source models with strong cross-language generalization.

---

### BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models

Authors: Yuanhao Li, Hongbo Wang, Xiaotang Shang, Xunzhu Tang, Yiming Cao, Xuhong Chen
Published: 2026-05-09
Categories: cs.AI, cs.SE
arXiv: https://arxiv.org/abs/2605.09134v3

Reinforcement learning for program repair is hindered by sparse execution feedback and coarse sequence-level rewards that obscure which edits actually fix bugs. We present BoostAPR, a three-stage framework addressing these challenges: (1) supervised fine-tuning on execution-verified demonstrations with reasoning traces, (2) training dual reward models--a sequence-level assessor and a line-level credit allocator--from execution outcomes, and (3) PPO optimization where the line-level model redistributes rewards to critical edit regions. This line-level credit assignment operates at an intermediate granularity naturally suited to code changes. Trained on SWE-Gym and evaluated on four benchmarks, BoostAPR achieves 40.7% on SWE-bench Verified (+22.9pp over base model), 24.8% on Defects4J (Python-to-Java transfer), 84.5% on HumanEval-Java, and 95.0% on QuixBugs, achieving competitive results among open-source models with strong cross-language generalization.

---

### KEPIL: Knowledge-Enhanced Prompt-Image Learning for Prompt-Robust Disease Detection

Authors: Haozhe Luo, Shelley Zixin Shu, Ziyu Zhou, Robert Berke, Mauricio Reyes
Published: 2026-05-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2605.09132v1

Vision--language models (VLMs) show promise for clinical decision support in radiology because they enable joint reasoning over radiological images and clinical text, thereby leveraging complementary clinical information. However, radiological findings are long-tailed in practice, leaving some conditions underrepresented and making zero-shot inference essential. Yet current CLIP-style medical VLMs are sensitive to prompt variations and often lack trustworthy external knowledge at inference time, which hinders reliable clinical deployment. We present \textit{KEPIL}, a prompt-robust framework that integrates curated medical knowledge to stabilize zero-shot generalization. KEPIL comprises: (i) \emph{dynamic prompt enrichment} using ontologies with LLM assistance, (ii) a \emph{semantic-aware contrastive loss} aligning embeddings of equivalent prompt variants via a dual-embedding objective, and (iii) \emph{entity-centric report standardization} to yield ontology-aligned representations. Across seven benchmarks, KEPIL achieves state-of-the-art zero-shot inference performance; under prompt-variation tests, it improves AUC by \(6.37\%\) on \textit{CheXpert} and by \(4.11\%\) on average. These results suggest that structured knowledge and robust prompt design are key to clinically reliable radiology-facing VLMs. Code will be released at https://github.com/Roypic/KEPIL.

---

### MCP-Cosmos: World Model-Augmented Agents for Complex Task Execution in MCP Environments

Authors: Giridhar Ganapavarapu, Dhaval Patel
Published: 2026-05-09
Categories: cs.AI, cs.MA
arXiv: https://arxiv.org/abs/2605.09131v1

The Model Context Protocol (MCP) has unified the interface between Large Language Models (LLMs) and external tools, yet a fundamental gap remains in how agents conceptualize the environments within which they operate. Current paradigms are bifurcated: Task-level planning often ignores execution-time dynamics, while reactive execution lacks long-horizon foresight. We present MCP-Cosmos, a framework that infuses generative World Models (WM) into the MCP ecosystem to enable predictive task automation. By unifying three disparate technologies, namely MCP, World Model, and Agent, we demonstrate that a "Bring Your Own World Model" (BYOWM) strategy allows agents to simulate state transitions and refine plans in a latent space before execution. We conducted experiments using two strategies, namely ReAct and SPIRAL with 2 planning models and 3 representative world models over 20+ MCP-Bench tasks. We observed improvements in Agent's environment interaction KPI such as tool success rate and tool parameter accuracy. The framework also offers new metrics such as Execution Quality to generate new insights about the effectiveness of world models compared to baseline.

---

### Data-driven Circuit Discovery for Interpretability of Language Models

Authors: Daking Rai, Mor Geva, Ziyu Yao
Published: 2026-05-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2605.09129v1

Circuit discovery aims to explain how language models (LMs) implement a specific task by localizing and interpreting a circuit, a computational subgraph responsible for the LM's behavior. Existing circuit discovery methods are hypothesis-driven; they first informally define a task with a dataset, and then apply a circuit discovery algorithm over that dataset to obtain a single circuit. This imposes two strong assumptions: that the LM implements the task with a single circuit, and that the dataset adequately represents the task as humans understand it. We systematically test these assumptions across four previously studied tasks and find that even minor dataset variations that preserve task semantics can produce circuits with low edge overlap and cross-dataset faithfulness. More strikingly, when applied to a mixed dataset with two distinct tasks whose separately discovered circuits have near-zero cross-faithfulness, existing methods still return a single circuit with high faithfulness across both tasks. This indicates that current methods discover dataset-specific circuits, rather than general task circuits. We propose Data-driven Circuit Discovery (DCD), a new discovery framework that drops both assumptions: instead of returning a single circuit for a dataset, DCD first clusters examples in the dataset by how similarly the model processes them and discovers a separate circuit for each group. This allows distinct mechanisms to appear separately rather than merged into a single circuit; each circuit explains its group, not the full task. Experiments show that DCD discovers multiple circuits per dataset, each more faithful to its group than a single circuit discovered by existing methods. Broadly, DCD lets the data reveal mechanistic structure within LMs, rather than relying on human-defined task boundaries that may not align with how models organize their computation.

---

### Internal vs. External: Comparing Deliberation and Evolution for Multi-Agent Constitutional Design

Authors: Hershraj Niranjani, Ujwal Kumar, Phan Xuan Tan
Published: 2026-05-09
Categories: cs.MA, cs.AI
arXiv: https://arxiv.org/abs/2605.09128v1

Multi-agent AI systems need behavioral constitutions, but it is unresolved whether such rules should emerge internally through agent self-governance or be discovered externally through optimization. We present the first controlled comparison of internal deliberation and external evolution across three social environments: a coordination grid-world, an iterated public goods game, and a bilateral trading market. Across 180 simulation runs, evolution significantly outperforms deliberation in collective-action settings (p < 0.01), while neither method improves outcomes in bilateral trading. A multiplier ablation reveals that evolution's advantage inverts when incentives shift: at pool multiplier (m = 0.75) the evolved constitution forces value-destroying cooperation and becomes the worst-performing method. Notably, no deliberation run across thirty trials ever proposed punishment -- the canonical cooperation-sustaining mechanism evolution reliably discovers -- suggesting external optimization wins on peaks while internal self-governance trades peaks for structural responsiveness.

---

### Cosine-Gated Adam-Decay: Drop-In Staleness-Aware Outer Optimization for Decoupled DiLoCo

Authors: Vatsal Shah, Jiahao Sun
Published: 2026-05-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2605.09126v1

Asynchronous DiLoCo systems may receive pseudo-gradients computed several outer rounds earlier, yet the standard Nesterov outer optimizer does not explicitly condition its update on per-update age. This can make the outer momentum buffer brittle under large controlled delays. We propose Cosine Gated Adam Decay (CGAD), a simple, drop-in, age-aware outer optimizer that scales each incoming pseudo-gradient by $σ(τ) = γ(τ) e^{-ατ}$ before it enters Adam's first- and second-moment buffers; the exponential models information decay and the cosine gate $γ(τ)$ smoothly zeroes contributions past a chosen cutoff. CGAD reduces to plain Adam at $τ=0$, adds two hyperparameters whose defaults transfer across scales, and extends to partial-sync schedulers via a per-fragment age-aware variant (PA-CGAD). For an idealized gated-adaptive update on smooth non convex objectives, we prove a non-asymptotic convergence bound whose staleness-bias term depends on $α$ alone, rather than on the realized maximum delay $τ_{\max}$; standard analyses of asynchronous momentum-SGD instead carry a $τ_{\max}^2$ factor. Empirically, on Llama style language model pretraining at 25M, 1B, and 7B parameters, CGAD trains stably across the controlled delays we sweep. The cosine cutoff acts as scale insurance: the closest baseline, Adam Decay (CGAD without the cutoff), is competitive at 25M but its seed-to-seed $σ$ at $τ=8$ grows 27x from 25M to 7B, pushing its single-shot risk (mean + $σ$) above the chance-level loss while CGAD's stays well below. The published Nesterov recipe is the least stable method on the full sweep.

---

### Transfer Learning of Multiobjective Indirect Low-Thrust Trajectories Using Diffusion Models and Markov Chain Monte Carlo

Authors: Jannik Graebner, Ryne Beeson
Published: 2026-05-09
Categories: eess.SY, cs.LG, math.OC
arXiv: https://arxiv.org/abs/2605.09125v1

Preliminary low-thrust spacecraft mission design is a global search problem characterized by a complex solution landscape, multiple objectives, and numerous local minima. During this phase, mission parameters are often not yet fully defined, requiring new solutions to be generated at a high cadence across varying parameter values. When combined with the indirect approach to optimal control, diffusion models can accelerate this search by learning distributions that represent high-quality initial costates. However, generating training data remains expensive, and opportunities exist to better exploit past data. We propose a transfer-learning framework that combines homotopy in a mission parameter with Markov chain Monte Carlo (MCMC) to generate training data more efficiently. The approach reformulates a multiobjective optimization problem as sampling from an unnormalized target distribution in costate space. We compare three MCMC algorithms on a planar multi-revolution transfer in the circular restricted three-body problem, with homotopy in the system mass parameter. The results show that gradient-based MCMC variants achieve the best trade-off between sample quality and computational cost. For the test transfer, the proposed framework generates 40 % more feasible solutions and achieves a higher-quality Pareto front than a state-of-the-art indirect approach based on adjoint control transformations and gradient-based optimization. Finally, the MCMC-generated samples are used to fine-tune a diffusion model conditioned on the mass parameter, enabling it to learn a global representation of the underlying solution distribution and efficiently generate new solutions. These findings establish the transfer-learning framework as a practical method for efficiently solving indirect trajectory optimization problems with varying parameters.

---

### A Communication-Theoretic Framework for LLM Agents: Cost-Aware Adaptive Reliability

Authors: Hamed Omidvar, Vahideh Akhlaghi
Published: 2026-05-09
Categories: cs.LG, cs.AI, cs.CL, cs.IT
arXiv: https://arxiv.org/abs/2605.09121v1

Agents built on large language models (LLMs) rely on a range of reliability techniques, including retry, majority voting, and self-consistency, that have been developed in parallel rather than within a common analytical framework. We observe that an LLM sampled at temperature $T$ is a discrete stochastic channel $p(y \mid x)$ in the sense of Shannon's coding theory, and use this identity as the entry point for such a framework grounded in communication theory. Each of these techniques is a special case of one of six classical reliability operators: diversity combining, hybrid retransmission, iterative generator-critic decoding, rateless sampling, structured redundant verification, and difficulty-adaptive routing. Within the framework we give two closed-form results: a noise-variance threshold above which uniform averaging beats quality-weighted averaging, and a contractivity criterion for generator-critic refinement, consistent with a contractive-to-divergent transition we observe between 3B- and 14B-parameter models. We further introduce a cost-aware semantic-nearest-neighbor router whose single Lagrangian knob traverses the quality-cost frontier without retraining. Across six channel configurations spanning local and cloud models on 69 hard tasks, no fixed model-technique-budget choice dominates, motivating per-task allocation. On a 300-item hard split of MMLU, GSM8K, and HumanEval, our router occupies the full empirical Pareto frontier: at matched quality, its normalized cost is ${\approx}56$\% lower than the strongest fixed technique; at matched normalized cost, it improves quality by ${\approx}7$\% ($26$\% over single-shot decoding). These results argue for consolidating these reliability techniques into a single tunable layer informed by channel coding.

---

### Personalized Alignment Revisited: The Necessity and Sufficiency of User Diversity

Authors: Enoch Hyunwook Kang
Published: 2026-05-09
Categories: cs.LG, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2605.09119v1

Personalized alignment aims to adapt large language models to heterogeneous user preferences, yet the precise theoretical conditions for its statistical efficiency have not been formally established. This paper characterizes the conditions under which personalized alignment achieves O(1) online regret and log(1/epsilon) offline sample complexity. We show that these optimal rates depend on a specific user-diversity condition: the population of user-specific heads must span the latent reward directions that can alter the optimal response. We prove that this condition is both necessary and sufficient. When it holds, simple greedy algorithms achieve benchmark efficiency; when it fails, every learner in a natural admissible class incurs at least logarithmic regret. Our results identify user diversity as the fundamental driver of personalized identifiability.

---

### Quantum Transfer Learning Shows Improved Robustness in Low-Data Regimes

Authors: Li-An Lo, Li-Yi Hsu, Hsien-Yi Hsieh
Published: 2026-05-09
Categories: quant-ph, cs.LG
arXiv: https://arxiv.org/abs/2605.09118v1

Transfer learning under limited data is a challenging setting, where models must adapt to new tasks with minimal supervision. Prior work has primarily focused on improving absolute accuracy in transfer learning. However, empirical evidence comparing quantum and classical models in realistic transfer learning settings remains limited, especially in low-data regimes. In this work, we systematically study the robustness of quantum models under reduced training data. We evaluate multiple quantum and classical architectures across diverse transfer tasks and retraining configurations, and quantify robustness using accuracy degradation and relative performance retention (RPR). Our results show that, although classical models often achieve higher peak performance, they exhibit significantly larger degradation when training data is limited. In contrast, quantum models maintain more stable performance across data regimes, indicating improved robustness and data efficiency. These findings provide empirical evidence that quantum models can offer improved robustness in low-resource transfer learning scenarios.

---

### Fit CATE Once: Model-Assisted Randomization Tests Without Sample Splitting

Authors: Fangnan Zheng, Yao Zhang
Published: 2026-05-09
Categories: stat.ME, stat.AP, stat.ML
arXiv: https://arxiv.org/abs/2605.09116v1

Randomization tests and flexible treatment-effect models offer complementary strengths for analyzing data from randomized panel experiments: the former provide valid inference under the known assignment mechanism, while the latter can capture complex patterns of effect heterogeneity. We develop model-assisted randomization tests that combine these strengths without sample splitting. The key idea is to estimate an unsigned version of the conditional average treatment effect (CATE) from the covariance structure of residualized outcomes, while leaving the realized assignments for randomization inference. The remaining sign can be chosen to best fit the observed outcomes. We establish identification and consistency for the proposed unsigned CATE estimators, as well as validity for the CATE-assisted randomization tests. Across synthetic and semi-synthetic experiments, the CATE-assisted randomization tests control Type I error and achieve higher power than covariate-adjusted and sample-split alternatives. Finally, we show that the assignment-free CATE estimates can be used to discover heterogeneous subgroups and test subgroup-specific treatment effects.

---

### AI Native Asset Intelligence

Authors: Gal Engelberg, Leon Goldberg, Konstantin Koutsyi, Boris Plotnikov, Tiltan Gilat, Ben Benhemo
Published: 2026-05-09
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2605.09115v2

Modern security environments generate fragmented signals across cloud resources, identities, configurations, and third-party security tools. Although AI-native security assistants improve access to this data, they remain largely reactive: users must ask the right questions and interpret disconnected findings. This does not scale in enterprise environments, where signal importance depends on exposure, exploitability, dependencies, and business context. Repeated AI queries may therefore produce unstable prioritization without a structured basis for comparing assets. This paper introduces AI-native asset intelligence, a framework that transforms heterogeneous security data into a structured intelligence layer for consistent, contextual, and proactive asset-level reasoning. The framework combines a modeling layer, representing assets, identities, relationships, controls, attack vectors, and blast-radius patterns, with a scoring layer that converts fragmented signals into a normalized measure of asset importance. The scoring system separates intrinsic exposure, based on misconfigurations and attack-vector evidence, from contextual importance, based on anomaly, blast radius, business criticality, and data criticality. AI contextualization refines severity and business/data classifications, while deterministic aggregation preserves consistency. We evaluate the scoring system on a production snapshot with 131,625 resources across 15 vendors and 178 asset types. Sensitivity analyses and ablations show that severity mappings control finding sensitivity, AI severity adjustment refines prioritization, attack-vector scoring responds to rare exploitability evidence, and contextual modulation selectively modifies exposed resources based on business or data importance. The results support AI-native asset intelligence as a foundation for stable prioritization and proactive security-posture reasoning.

---

### Contextual Plackett-Luce: An Efficient Neural Model for Probabilistic Sequence Selection under Ambiguity

Authors: Noam Mizrachi, Nadav Har-Tuv, Shai Shalev-Shwartz
Published: 2026-05-09
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2605.09112v1

Selecting a coherent sequence or subset of elements is a fundamental problem in structured prediction, arising in tasks such as detection, trajectory forecasting, and representative subset selection. In many such settings, the target is inherently ambiguous: each input admits multiple valid outputs, while supervision provides only a single sampled instance. This induces a mismatch between the underlying multi-modal target distribution and the observed training signal. We propose Contextual Plackett-Luce (CPL), a structured probabilistic model for sequence selection that extends the classical Plackett-Luce model to a context-dependent setting following an Ising-style parameterization with unary and pairwise interaction terms. CPL can be viewed as a hybrid between fully autoregressive prediction and parallel sequence selection: autoregressive models effectively capture uncertainty but are computationally expensive on modern parallel hardware such as GPUs, while parallel methods are efficient but struggle to represent multi-modal dependencies. CPL combines the strengths of both by constructing the parameters of a probabilistic selection model in a fully parallel manner, followed by a lightweight autoregressive selection process in which each step applies incremental updates to contextual logits. This decoupling of parallel scoring and sequential selection enables efficient computation without sacrificing expressivity. We evaluate CPL on two structured selection tasks: multi-modal path prediction and representative subset selection. CPL achieves improved structural consistency and robustness under ambiguous supervision compared to strong parallel baselines.

---

### When (and How) to Trust the Expert: Diagnosing Query-Time Expert-Guided Reinforcement Learning

Authors: Yann Berthelot, Philippe Preux, Riad Akrour
Published: 2026-05-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2605.09109v1

Many continuous-control problems ship with a competent but suboptimal controller (a tuned PID, a hand-designed gait). A growing family of methods uses such controllers as queryable experts during RL, but each method has been proposed in isolation, on a different benchmark, without imperfect-expert testing. We harmonize the comparison on a shared SAC backbone, common HPO and evaluation protocols, 100/50 seeds per (env, method), and a degradation sweep over expert undertuning, action bias, and observation noise. The comparison surfaces three failure modes single-paper evaluations miss: (F1) a critic blind spot under argmax-plus-bootstrap that drags IBRL below no-expert SAC on experts close to the no-expert-RL ceiling (RL-near-ceiling, distinct from the absolute physical ceiling); (F2) residual saturation on far-from-optimal experts; and (F3) warm-start buffer poisoning that collapses training-time-handoff methods under deployment-time expert undertuning. No single method dominates: each wins on one task-structure regime and fails predictably elsewhere; on RL-near-ceiling experts (FourTank, GlassFurnace) no query-time method clears the expert within our 1M-step budget, leaving open whether this is a fundamental wall or a budget effect. We convert the spread into a testable decision rule keyed on three pre-training observables (expert quality, task termination, perturbation type). The benchmark, taxonomy, and decision rule are the primary contribution; we additionally describe EDGE, a softmax-over-ensemble-LCB design point used to demonstrate that both axes the taxonomy points to (gate form, scoring rule) are individually exploitable.

---

### Fin-Bias: Comprehensive Evaluation for LLM Decision-Making under human bias in Finance Domain

Authors: Xiaoyu Hu, Jinman Zhao
Published: 2026-05-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2605.09106v1

Large language models (LLMs) are increasingly deployed in financial contexts, raising critical concerns about reliability, alignment, and susceptibility to adversarial manipulation. While prior finance-related benchmarks assess LLMs' capabilities in stock trading, they are often restricted to small sample and fail to demonstrate LLM susceptibility to context with potential human bias. We introduce Fin-Bias (financial herding under long and uncertain financial context), a benchmark for evaluating LLM investment decision-making when faced with uncertainty and possible human-biased opinions. Fin-Bias includes 8868 long firm-specific analyst reports, including firm aspects summarized and analyzed by sophisticated analysts with investment ratings (Bullish/Neutral/Bearish) spanning from various industries. We present large language models with firm analyst reports with/without analyst investment ratings and even with 'fake' rating, to get investment ratings generated by LLMs. Our results reveal that LLMs tend to herd the explicit bias in context. We also develop a method to detect potential human opinions, which can encourage LLMs to think independently, some models even exceed human performance in predicting future stock return.

---

### Token Economics for LLM Agents: A Dual-View Study from Computing and Economics

Authors: Yuxi Chen, Junming Chen, Chenyu He, Yiwei Li, Yicheng Ji, Yifan Wu, Dingyu Yang, Lansong Diao, Lidan Shou, Hongliang Zhang, Huan Li, Gang Chen
Published: 2026-05-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2605.09104v1

As LLM agents evolve, tokens have emerged as the core economic primitives of Agentic AI. However, their exponential consumption introduces severe computational, collaborative, and security bottlenecks. Current surveys remain fragmented across system optimization, architecture design, and trust, lacking a unified framework to evaluate the fundamental trade-off between output quality and economic cost. To bridge this gap, this survey presents the first comprehensive survey of Token Economics. By unifying computer science and economics, we conceptualize tokens as production factors, exchange mediums, and units of account. We synthesize existing literature across a four-dimensional taxonomy: (1) Micro-level (Single Agent): Optimizing budget-constrained factor substitution via neoclassical firm theory. (2) Meso-level (Multi-Agent Systems): Minimizing collaboration friction using transaction cost and principal-agent theories. (3) Macro-level (Agent Ecosystems): Addressing congestion externalities and pricing via mechanism design. (4) Security: Internalizing adversarial threats as endogenous economic constraints. Finally, we outline frontier directions, including differentiable token budgets and dynamic markets, to lay the theoretical foundation for scalable next-generation agent systems.

---

### GRC: Unifying Reasoning-Driven Generation, Retrieval and Compression

Authors: Zhongtao Miao, Qiyu Wu, Yoshimasa Tsuruoka
Published: 2026-05-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2605.09100v2

Text embedding and generative tasks are usually trained separately based on large language models (LLMs) nowadays. This causes a large amount of training cost and deployment effort. Context compression is also a challenging and pressing task, which is vital to reasoning-driven generation, and agentic tasks requiring long context and continual learning. In this paper, we explore how to unify reasoning-driven generation, reasoning-enhanced text representation and context compression tasks in one forward pass for LLMs. Through meta latent tokens and a unified generative, representative and compressive tuning approach, we propose a training framework named GRC that bridges the three tasks. The trained models can accomplish three objectives in a single forward pass while maintaining modular, LEGO-style flexibility during inference. This design greatly reduces the deployment effort for retrieval-augmented generation (RAG) and achieves efficient inference and three times data utilization during training. Furthermore, this framework design enables a new paradigm for text embedding: self-reason-latent embeds, and a new generation paradigm, latent memory-augmented generation, where compressed and internalized KV cache with O(1) length is used as the updatable memory. We also propose hybrid paged attention to speed up the inference of our models. Extensive experiments on reasoning-intensive retrieval benchmarks, generative tasks, document compression, latency evaluation, and RAG settings demonstrate the effectiveness of our method and may shed light on the truly unified model that can handle reasoning-driven generation, embedding and compression tasks seamlessly.

---

### Dynamic Meta-Metrics: Source-Sentence Conditioned Weighting for MT Evaluation

Authors: Luke Zhang, Justin Vasselli, Aditya Khan, York Hay Ng, En-Shiun Annie Lee
Published: 2026-05-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2605.09098v2

We propose Dynamic Meta-Metrics (DMM), a framework for machine translation evaluation that learns source-sentence conditioned combinations of existing metrics. Rather than relying on a single static ensemble or language-specific weighting, DMM adapts the metric combination based on properties of the source segment. We study hard conditioning, which fits an interpretable combiner per cluster, and an exploratory soft-conditioned extension whose weights vary continuously with source-cluster responsibilities. We evaluate DMM on the WMT Metrics Shared Task data across multiple language pairs using pairwise agreement measures at the system and segment levels. Across settings, MLP-based combinations outperform linear and Gaussian process-based ensembles, and introducing soft conditioning yields gains over linear models.

---

### Bridging Spectral Operator Learning and U-Net Hierarchies: SpectraNet for Stable Autoregressive PDE Surrogates

Authors: Enrique Hernández Noguera, Md Meftahul Ferdaus, Elias Ioup, Mahdi Abdelguerfi, Julian Simeonov
Published: 2026-05-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2605.09096v1

Neural operators for time-dependent PDEs face a structural tension: spectral architectures (FNO and descendants) inherit exponential rollout-error growth from their one-step Lipschitz constant, while hierarchical U-Net operators trade resolution invariance for multi-scale detail. We introduce SpectraNet, an autoregressive neural operator that composes truncated spectral convolutions inside a U-Net hierarchy with a Residual-Target Spectral Block trained under a Semigroup-Consistency Loss. The residual-target parametrization replaces L^T stability blow-up with linear T*delta drift, and the spectral path's parameter count is Theta(L w^2 M^2), independent of grid N. Under a single unified protocol against 16 published neural-operator baselines on Navier-Stokes nu=1e-5 at 64x64, SpectraNet reaches test relative L2 = 0.0822 at 2.04M parameters -- 2.33x fewer than canonical FNO at ~20% lower error -- and wins five of six rows in a cross-PDE comparison against FNO (NS at nu in {1e-4, 1e-3}, PDEBench Shallow-Water 2D and Diffusion-Reaction, with the Active-Matter row going to FNO inside its seed spread). Trained from scratch at native 128^2 under the same protocol, SpectraNet improves to 0.0724 while FNO regresses to 0.3080. Free rollout stays bounded for T=100 where FNO diverges across all 200 test trajectories. On consumer CPU at B=1, SpectraNet runs sub-200ms while the full-attention Transformer that wins raw L2 pays ~60x latency; we do not claim to beat that Transformer on raw L2, only to dominate the lightweight (<=5M parameter, sub-200ms CPU) Pareto frontier. Source code: https://github.com/Enrikkk/spectranet

---

### A Tale of Two Problems: Multi-Task Bilevel Learning Meets Equality Constrained Multi-Objective Optimization

Authors: Zhiyao Zhang, Myeung Suk Oh, Zhen Qin, Jiaxiang Li, Xin Zhang, Jia Liu
Published: 2026-05-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2605.09094v2

In recent years, bilevel optimization (BLO) has attracted significant attention for its broad applications in machine learning. However, most existing works on BLO remain confined to the single-task setting and rely on the lower-level strong convexity assumption, which significantly restricts their applicability to modern machine learning problems of growing complexity. In this paper, we make the first attempt to extend BLO to the multi-task setting under a relaxed lower-level general convexity (LLGC) assumption. To this end, we reformulate the multi-task bilevel learning (MTBL) problem with LLGC into an equality constrained multi-objective optimization (ECMO) problem. However, ECMO itself is a new problem that has not yet been studied in the literature. To address this gap, we first establish a new Karush-Kuhn-Tucker (KKT)-based Pareto stationarity as the convergence criterion for ECMO algorithm design. Based on this foundation, we propose a weighted Chebyshev (WC)-penalty algorithm that achieves a finite-time convergence rate of $O(ST^{-\frac{1}{2})$ to KKT-based Pareto stationarity in both deterministic and stochastic settings, where $S$ denotes the number of objectives, and $T$ is the total iterations. Moreover, by varying the preference vector over the $S$-dimensional simplex, our WC-penalty method systematically explores the Pareto front. Finally, solutions to the ECMO problem translate directly into solutions for the original MTBL problem, thereby closing the loop between these two foundational optimization frameworks.

---

### Character-Level Transformer for Tajik-Persian Transliteration with a Parallel Lexical Corpus

Authors: Mullosharaf K. Arabov
Published: 2026-05-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2605.09092v1

This study addresses automatic transliteration from Tajik (Cyrillic script) to Persian (Perso-Arabic script). We present a curated, lexicographically verified parallel corpus of 52,152 Tajik--Persian words and short phrases, compiled from printed dictionaries, encyclopedic sources, and manually verified online resources. To the best of our knowledge, this is one of the largest publicly available word-level corpora for Tajik--Persian transliteration. Using this corpus, we train a character-level sequence-to-sequence Transformer model and evaluate it using Character Error Rate (CER) and exact-match accuracy. The Transformer achieves a CER of 0.3216 and an exact-match accuracy of 0.3133, outperforming both dictionary-based rule-based and recurrent neural baselines. With beam search (k=3), performance further improves to CER 0.3182 and accuracy 0.3215. We describe the data collection and preprocessing pipeline, model architecture, and experimental protocol, and report a part-of-speech analysis showing performance differences across lexical categories. All preprocessing scripts, deterministic splits into training, validation, and test sets, and training configurations are released to support reproducibility and further research on Tajik and related Persian dialects. The corpus supports research in character-level transliteration, cross-script NLP, and lexicographic applications.

---

### Investigating Anisotropy in Visual Grounding under Controlled Counterfactual Perturbations

Authors: Gabriele Lombardo, Luigi Maiorana, Liliana Lo Presti, Marco La Cascia
Published: 2026-05-09
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2605.09090v1

Visual Grounding benchmarks assume that the object described by a referring expression is always present in the image, and grounding models are therefore rarely evaluated under semantically mismatched captions. In such cases, models frequently exhibit approximation behavior, producing a plausible bounding box that satisfies only part of the expression (\eg, preserving the original object while ignoring modified contextual cues). Because mismatched captions represent realistic edge cases, this behavior compromises reliability and raises concerns from an explainability perspective. Identifying its underlying causes is thus essential for improving model faithfulness and interpretability. Adopting a mechanistic interpretability viewpoint, this work examines whether embedding anisotropy contributes to counterfactual failures. A similarity-controlled counterfactual caption generation protocol is introduced to systematically perturb object or contextual components within predefined embedding similarity intervals, enabling a fine-grained analysis of grounding behavior as a function of alignment. Experiments on two Transformer-based models with markedly different embedding geometries (BERT-based TransVG and CLIP-based SwimVG) reveal no meaningful correlation between cosine similarity and approximation. These findings suggest that anisotropy alone does not account for counterfactual errors, and that robustness requires investigating finer-grained geometric properties of the embedding space.

---

### Field-Localized Forgery Detection for Digital Identity Documents

Authors: Abhishek Kumar, Riya Tapwal, Carsten Maple, Mark Hooper
Published: 2026-05-09
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2605.09089v2

Digital onboarding and eKYC systems used by banks, fintech platforms, telecom providers, and other third-party services commonly verify users by comparing an uploaded identity document with a selfie or live facial capture. This workflow is convenient, but it also makes verification systems vulnerable to localised document manipulations, such as replacing the facial photograph, editing textual identity fields, or altering both. Existing image-forgery detectors are largely designed for natural images and do not explicitly account for the structured layout of identity documents, where security-relevant information is concentrated in specific semantic fields. We propose FLiD, a lightweight field-localised framework for identity-document forgery detection. Instead of processing the full document image, FLiD localises the facial and textual regions using a fine-tuned YOLO11 detector, extracts compact representations with a frozen MobileNetV3-Small backbone, and classifies field-level forgeries using a small 191K-parameter head. Face and text detectors are trained independently and combined through score-level fusion for documents containing simultaneous manipulations. Under 5-fold cross-validation, FLiD achieves AUC scores of 0.834, 0.926, and 0.837 for face, text, and combined attacks, respectively, while reducing Equal Error Rate by 28--29 percentage points compared with a full-document baseline. FLiD also outperforms general-purpose forgery detectors while requiring 13x fewer trainable parameters and 21x fewer FLOPs per field (~8x per full multi-field document), making it a promising candidate for data and resource constrained KYC deployment.

---

### Field-Localized Forgery Detection for Digital Identity Documents

Authors: Abhishek Kumar, Riya Tapwal, Carsten Maple, Mark Hooper
Published: 2026-05-09
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2605.09089v1

Digital identity verification systems used in remote onboarding rely on document images to authenticate users, making them vulnerable to localized manipulations of key identity fields such as facial photographs and textual information. Existing forgery detection methods, developed primarily for natural-image forensics, show limited transferability to structured identity documents. We propose FLiD, a lightweight field-localized framework that targets critical identity regions rather than processing full-document images. A fine-tuned object detector first localizes face and text fields; a frozen MobileNetV3-Small backbone then extracts compact field-level embeddings, which are classified by lightweight neural network with only 191K trainable parameters. FLiD achieves AUC scores of 0.880 (face), 0.954 (text), and 0.923 (both-field attacks), with corresponding EERs of 18.05%, 11.61%, and 15.16%, representing absolute reductions of 29-35 percentage points over a full-document baseline trained from scratch. FLiD also consistently outperforms general-purpose manipulation detectors (TruFor, MMFusion, UniVAD) across all attack scenarios while requiring 13x fewer parameters and 21x fewer FLOPs

---

### Towards Trustworthy Audio Deepfake Detection: A Systematic Framework for Diagnosing and Mitigating Gender Bias

Authors: Aishwarya Fursule, Shruti Kshirsagar, Anderson R. Avila
Published: 2026-05-09
Categories: cs.SD, cs.LG
arXiv: https://arxiv.org/abs/2605.09087v1

Audio deepfake detection systems are increasingly deployed in high-stakes security applications, yet their fairness across demographic groups remains critically underexamined. Prior work measures gender disparity but does not investigate where it comes from or how to fix it systematically. We present the first diagnosis-first framework that identifies bias source before applying targeted mitigation, evaluated on two models, AASIST and Wav2Vec2+ResNet18, on ASVSpoof5. Our diagnosis shows that bias does not stem from imbalanced training data but from acoustic representation differences, gender leakage in learned features, and structural evaluation asymmetry. We test mitigation strategies across in-processing, post-processing and combined families, including novel methods introduced in this work. Adjusting the decision threshold separately per gender reduces unfairness by 54% to 75% at no cost to detection accuracy, and our new epoch-level fairness regularisation method outperforms existing per-batch approaches. Adversarial debiasing succeeds only when gender leakage is localised, and fails when it is diffuse, an outcome correctly predicted by our diagnosis before training. No single method fully closes the fairness gap, confirming that bias sources must be identified before fixes are applied and that fairer benchmark design is equally important

---

### Constant-Target Energy Matching: A Unified Framework for Continuous and Discrete Density Estimation

Authors: Zhijun Zeng, Yixuan Jiang, Pipi Hu, Zuoqiang Shi
Published: 2026-05-09
Categories: cs.AI, math.PR
arXiv: https://arxiv.org/abs/2605.09085v1

Density estimation is a central primitive in probabilistic modeling, yet continuous, discrete, and mixed-variable domains are often treated by separate objectives, limiting the ability to exploit a common statistical structure across data types. Continuous score-based methods rely on log-density gradients, while discrete extensions typically use concrete score whose unbounded targets become unstable near low-probability states. We introduce Constant-Target Energy Matching (CTEM), a unified energy-based framework for density estimation on general state spaces. CTEM replaces ordinary density-ratio regression with a bounded energy-difference transform and derives from it a sample-only training objective with the constant target 1. The learned scalar potential recovers log p without partition-function estimation or explicit unbounded ratio regression. Across continuous, discrete, and mixed-variable benchmarks, CTEM substantially improves density estimation over competitive baselines and yields higher-quality samples under standard sampling procedures.

---

### FactoryNet: A Large-Scale Dataset toward Industrial Time-Series Foundation Models

Authors: Karim Othman, Jonas Petersen, Matei Ignuta-Ciuncanu, Camilla Mazzoleni, Federico Martelli, Alessandro Lombardi, Riccardo Maggioni, Philipp Petersen
Published: 2026-05-09
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2605.09081v4

We introduce the first universal pretraining corpus for industrial time-series data: FactoryNet. 51M datapoints across 23k end-to-end task executions (13.3k real, 9.8k synthetic) on six embodiments, unified by a shared schema that enables robust zero-shot cross-embodiment transfer and highly parameter-efficient anomaly detection. We introduce a novel schema: Setpoint, Effort, Feedback, Context (S-E-F-C) underlying the whole pipeline that maps any actuated system into a common representational frame. The corpus spans 27 annotated anomaly types alongside healthy baselines and counterfactual pairs across robotic manipulation and machining domains. Cross-embodiment transfer experiments yield positive results: under bias-aware metrics our model demonstrates fair cross-embodiment transfer capabilities on the evaluated source-target pair, while 24 schema-aligned signals achieves competitive anomaly detection performance compared to high-dimensional baselines. We release FactoryNet as a growing, multi-embodiment dataset to drive progress toward industrial foundation models.

---

### CauSim: Scaling Causal Reasoning with Increasingly Complex Causal Simulators

Authors: Nicolás Astorga, Anita Kriz, Mihaela van der Schaar
Published: 2026-05-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2605.09079v1

Despite surpassing human performance across mathematics, coding, and other knowledge-intensive tasks, large language models (LLMs) continue to struggle with causal reasoning. A core obstacle is the target data itself: causal systems are complex and often expressed in non-executable forms, while ground-truth answers to causal queries are inherently scarce. We introduce CauSim, a framework that turns causal reasoning from a scarce-label problem into a scalable supervised one. CauSim constructs increasingly complex causal simulators: executable structural causal models (SCMs), incrementally built by LLMs, that scale to globally complex systems while maintaining verifiable answers to causal queries. CauSim operates across representations by formalizing non-executable causal knowledge into code, enabling data augmentation, and translating executable SCMs into natural language, enabling supervision in previously difficult-to-supervise representations. We structure our research into two parts: (1) how to construct increasingly complex causal simulators, and (2) a systematic study of what CauSim enables, demonstrating generalization across representations, consistent gains from curriculum scaling and data volume, LLM self-improvement through self-generated simulators, and data augmentation via formalization of existing domain knowledge.

---

### Robust Multi-Agent LLMs under Byzantine Faults

Authors: Haejoon Lee, Vincent-Daniel Yun, Dimitra Panagou, Sai Praneeth Karimireddy
Published: 2026-05-09
Categories: cs.MA, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2605.09076v3

Large language model (LLM) agents increasingly collaborate over peer-to-peer networks to improve their reliability. However, these same interactions can also introduce vulnerability to unreliable or Byzantine agents that can propagate incorrect information and degrade overall system performance. To address this, we propose Self-Anchored Consensus (SAC), a fully decentralized filter-and-refine protocol in which agents iteratively exchange responses, locally evaluate and filter unreliable messages, and refine their own outputs. We present $(F{+}1)$-robustness conditions on the communication graph that ensure honest agents preserve and propagate reliable information despite Byzantine influence. Experiments across diverse open- and closed-weight LLMs on mathematical and commonsense reasoning benchmarks show that SAC effectively suppresses Byzantine influence and consistently improves performance across diverse communication topologies, whereas prior methods degrade significantly under Byzantine attacks.

---

### Robust Multi-Agent LLMs under Byzantine Faults

Authors: Haejoon Lee, Vincent-Daniel Yun, Dimitra Panagou, Sai Praneeth Karimireddy
Published: 2026-05-09
Categories: cs.MA, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2605.09076v2

Large language model (LLM) agents increasingly collaborate over peer-to-peer networks to improve their reliability. However, these same interactions can also become a source of vulnerability, as unreliable or Byzantine agents may sway neighboring agents toward incorrect conclusions and degrade overall system performance. Existing methods rely on leader-based coordination or self-reported confidence, both of which are susceptible to adversarial manipulation. We study decentralized LLM multi-agent systems (LLM-MAS) and propose Self-Anchored Consensus (SAC), a fully decentralized iterative filter-and-refine protocol in which agents iteratively exchange responses, locally evaluate and filter unreliable messages, and refine their own outputs. We present $(F{+}1)$-robustness conditions for the communication graph that ensure honest agents preserve and propagate reliable information despite Byzantine influence. Experiments on mathematical and commonsense reasoning benchmarks show that SAC effectively suppresses Byzantine influence and consistently improves performance across diverse communication topologies, whereas prior methods degrade under adversarial conditions.

---

### Optimality of Sub-network Laplace Approximations: New Results and Methods

Authors: Swarnali Raha, Kshitij Khare, Rohit K Patra
Published: 2026-05-09
Categories: stat.ML, cs.LG
arXiv: https://arxiv.org/abs/2605.09075v2

Although the Laplace approximation offers a simple route to uncertainty quantification in deep neural networks, its reliance on inverting large Hessian matrices has motivated a range of computationally feasible low-dimensional or sparse approximations. A prominent class of such methods - sub-network Laplace approximations, constructs surrogates by restricting attention to a small subset of parameters. Existing approaches in this family typically rely on diagonal, layer-wise, or other architectural heuristics for subset selection, which ignore cross-parameter interactions and lack formal optimality guarantees. In this paper, we provide a rigorous theoretical analysis of the sub-network Laplace paradigm. We prove that all sub-network Laplace methods systematically underestimate the predictive variance of the full Laplace posterior, and that this bias decreases monotonically as the retained sub-matrix expands. Leveraging this insight, we propose two principled, analytically grounded sub-network Hessian approximations: \textit{Gradient-Laplace} selects parameters with the largest average squared gradients of the model output with respect to the parameters over a reference dataset; while \textit{Greedy-Laplace} iteratively refines this selection by accounting for off-diagonal interactions in the precision matrix. We establish theoretical guarantees characterizing their optimality properties and show that Gradient-Laplace provably outperforms existing heuristic approaches. Extensive numerical studies across diverse settings indicate that these methods perform strongly relative to existing benchmarks.

---

### Probability-Flow Distillation: Distribution Matching in Parameter Space

Authors: Rohith Ramanan, A. N. Rajagopalan
Published: 2026-05-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2605.09071v2

Score distillation methods use pretrained diffusion models as priors for optimizing parameters through differentiable forward models, most notably in text-to-3D generation. Yet the distribution they induce over those parameters is not well understood. Observing that existing distillation methods reduce to one of three: Score Distillation Sampling (SDS), Score Distillation via Inversion (SDI), and Variational Score Distillation (VSD), we extend the particle variational inference view of VSD to the other two. We show that SDS collapses onto the modes of the target, while SDI converges to a contracted version of it, and explain why SDI needs a negative classifier-free guidance scale. Next, we observe that the DDIM posterior mean equals a single Euler step of the probability-flow ODE (PF-ODE). Replacing this step in SDI with a full reverse solve makes the target a fixed point, but it requires solving two concatenated PF-ODEs. Dropping a Jacobian from the resulting gradient gives Probability-Flow Distillation (PFD), which requires solving only the forward PF-ODE. Experiments on synthetic targets, the CelebA dataset, and text-to-3D generation support our analysis and demonstrate the practical effectiveness of PFD.

---

### Probability-Flow Distillation: Exact Wasserstein Gradient Flow for High-Fidelity 3D Generation

Authors: Rohith Ramanan, A. N. Rajagopalan
Published: 2026-05-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2605.09071v1

Score Distillation Sampling (SDS) and its variants have been widely used for text-to-3D generation by distilling 2D image diffusion priors. However, the standard SDS objective is prone to severe mode collapse, frequently yielding over-smoothed and over-saturated results. Although recent advancements, such as Score Distillation via Inversion (SDI), mitigate these artifacts and produce visually sharper models, they ultimately fail to faithfully capture the full target distribution. In this work, we show that the bottleneck limiting the sampling capacity of SDI stems from its reliance on the posterior mean estimator, which is mathematically equivalent to a single-step Euler approximation of the deterministic reverse DDIM trajectory. To address this, we propose a naturally motivated extension termed Probability-Flow Distillation (PFD). We establish that PFD corresponds exactly to a Wasserstein gradient flow, thereby inducing principled distribution-matching dynamics. Finally, we show that PFD can synthesize 3D assets with fine-grained, high-fidelity details and achieve improved quality compared to existing methods.

---

### Single-Configuration Attack Success Rate Is Not Enough: Jailbreak Evaluations Should Report Distributional Attack Success

Authors: Carsten Maple, Abhishek Kumar, Riya Tapwal
Published: 2026-05-09
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2605.09070v1

Many jailbreak attack research papers report attack success rates for a limited number of parameter settings, even though there are many combinations of parameter settings that could be used. Further, when new jailbreak papers are released, they often benchmark results against single configurations of existing attacks. This position paper argues such practices are fundamentally insufficient for characterising the threat posed by parameterised jailbreak attacks, and comparing attacks. Most jailbreak attacks expose multiple internal parameters, system prompt templates, conversation rounds, cipher dispersion, teaching shots, and ASR varies substantially across these parameters. Reporting only the best-case configuration discards two pieces of information that defenders genuinely need: how typical that performance is across the variant space, and how much of the attack surface is missed by selecting a single variant. We propose two new measures for jailbreak attacks: the Variant Sensitivity Measure (VSM) and Union Coverage (UC). VSM quantifies how far the best reported ASR deviates from the mean ASR across the tested variant space, UC is the total fraction of prompts resulting in unsafe responses across all tested configurations. We empirically demonstrate the importance of these measures using two attack families across three open-source target models. For PAIR, the best template reaches 69% ASR on Mistral-7B and 75% on Qwen3-0.6B, while UC rises to 88% and 93%, respectively. For bijection on Mistral-7B, the best variant reaches 81% ASR, but the 36-variant union covers 100% of HarmBench-100 prompts. We argue that distributional reporting, publishing VSM alongside ASR and enumerating variant coverage as fully as compute allows, should become the new minimum standard for parameterised jailbreak evaluation.

---

### Reducing Annotation Burden for Femoral Cartilage Segmentation in Knee MRI via Cross-Sequence Transfer Learning

Authors: Francesco Chiumento, Gianluigi Crimi, Elisa Moretta, Rocco Milieri, Alberto Bazzocchi, Giulio Vara, Giacomo Dal Fabbro, Stefano Zaffagnini, Fulvia Taddei, Serena Bonaretti
Published: 2026-05-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2605.09067v1

Purpose: To develop and evaluate cross-sequence transfer learning for automatic femoral cartilage segmentation, testing bidirectional transfer between dual-echo steady-state (DESS) and sagittal proton density-weighted 3D fast spin-echo (Cube) sequences. Materials and Methods: We optimized a modified 2D U-Net on 507 DESS images from the Osteoarthritis Initiative (OAI). We then established same-sequence baselines using subject-level cross-validation on a subset of 44 OAI DESS images and 44 Cube images acquired at the Istituto Ortopedico Rizzoli, Bologna, Italy. Each subset included 22 non-lesioned and 22 lesioned subjects. Finally, we performed transfer learning across sequences by fine-tuning the pretrained models on the target sequence with increasing training set sizes to study convergence, while keeping validation and test sets fixed. Segmentations were evaluated using Dice similarity coefficient (DSC) and average surface distance (ASD). Lesion effects were assessed with two-sided Mann-Whitney U tests with Bonferroni correction. Results: Same-sequence training yielded higher accuracy on DESS than Cube (DSC, $0.900$ vs $0.830$; $P < .001$). Cube-to-DESS transfer matched DESS performance (DSC, $0.903 \pm 0.032$ vs $0.900 \pm 0.027$), reaching a performance plateau at 9 training subjects. DESS-to-Cube yielded a lower combined DSC ($0.802 \pm 0.049$ vs $0.830 \pm 0.042$), reaching a plateau at 24 training subjects. Lesions did not affect DESS ($P \ge .39$) but reduced Cube accuracy (DSC, $0.805$ vs $0.856$; $P < .001$). Conclusion: Transfer learning across sequences can substantially reduce target-sequence annotation requirements for femoral cartilage segmentation, but performance is direction- and sequence-dependent, and the effects of lesions on segmentation may vary across MRI sequences.

---

### Dependency-Aware Discrete Diffusion for Scene Graph Generation

Authors: Rajalaxmi Rajagopalan, Romit Roy Choudhury
Published: 2026-05-09
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2605.09065v1

Scene graphs (SGs) represent objects and their relationships as structured graphs, enabling applications in image generation, robotics, and 3D understanding. Recent work suggests that conditioning image generation on scene graphs improves compositional fidelity compared to text-only prompting. However, since users typically provide text rather than structured graphs, a key challenge is to generate scene graphs from natural language. Prior work on discrete diffusion has demonstrated success in generating generic graphs such as molecules and circuits, but fails to account for the hierarchical structure and strong dependencies between objects, edges, and relations in scene graphs. We address this limitation by introducing a dependency-aware, hierarchically constrained discrete diffusion model for scene graph generation. Our approach decouples structure and semantics across the forward and reverse processes, enabling the model to capture conditional dependencies. At inference time, we perform training-free conditioning to sample text-aligned scene graphs. We evaluate our method on standard SG benchmarks and demonstrate improvements over both continuous and discrete graph generation baselines across graph and layout metrics. When fed to downstream image generation, our approach yields improved compositional alignment compared to text-to-image models, particularly in multi-object scenarios.

---

### Soohak: A Mathematician-Curated Benchmark for Evaluating Research-level Math Capabilities of LLMs

Authors: Guijin Son, Seungone Kim, Catherine Arnett, Hyunwoo Ko, Hyein Lee, Hyeonah Kang, Jiang Longxi, Jin Yun, JungYup Lee, Kyungmin Lee, Sam Yoosuk Kim, Sang Park et al.
Published: 2026-05-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2605.09063v3

Following the recent achievement of gold-medal performance on the IMO by frontier LLMs, the community is searching for the next meaningful and challenging target for measuring LLM reasoning. Whereas olympiad-style problems measure step-by-step reasoning alone, research-level problems use such reasoning to advance the frontier of mathematical knowledge itself, emerging as a compelling alternative. Yet research-level math benchmarks remain scarce because such problems are difficult to source (e.g., Riemann Bench and FrontierMath-Tier 4 contain 25 and 50 problems, respectively). To support reliable evaluation of next-generation frontier models, we introduce Soohak, a 439-problem benchmark newly authored from scratch by 64 mathematicians. Soohak comprises two subsets. On the Challenge subset, frontier models including Gemini-3-Pro, GPT-5, and Claude-Opus-4.5 reach 30.4%, 26.4%, and 10.4% respectively, leaving substantial headroom, while leading open-weight models such as Qwen3-235B, GPT-OSS-120B, and Kimi-2.5 remain below 15%. Notably, beyond standard problem solving, Soohak introduces a refusal subset that probes a capability intrinsic to research mathematics: recognizing ill-posed problems and pausing rather than producing confident but unjustified answers. On this subset, no model exceeds 50%, identifying refusal as a new optimization target that current models do not directly address. To prevent contamination, the dataset will be publicly released in late 2026, with model evaluations available upon request in the interim.

---

### A Market-Rule-Informed Neural Network for Efficient Imbalance Electricity Price Forecasting

Authors: Runyao Yu, Julia Lin, Derek W. Bunn, Jochen Stiasny, Wentao Wang, Yujie Chen, Tara Esterl, Peter Palensky, Jochen L. Cremer
Published: 2026-05-09
Categories: q-fin.CP, cs.LG
arXiv: https://arxiv.org/abs/2605.09061v1

Accurate and efficient imbalance electricity price forecasting is critical for industrial energy trading systems, especially as battery assets and automated bidding pipelines increasingly participate in balancing markets. However, real-time forecasting is complicated by nonlinear market-rule-based price formation, heterogeneous input signals, and incomplete data availability caused by communication delays, publication lags, and measurement outages. This paper proposes a market-rule-informed neural forecasting framework that embeds imbalance price formation rules into the latent space of an expressive neural network. The proposed framework preserves raw signal information while exploiting transparent market-rule priors. We further analyze operational robustness by removing price-component information and characterize how forecasting performance scales with input length and forecasting horizon. Experimental results show that the proposed model achieves competitive forecasting performance with substantially fewer trainable parameters and shorter training time than generic deep learning baselines. Experimental results show that the proposed model achieves competitive forecasting performance with substantially fewer trainable parameters and shorter training time than generic deep learning baselines, demonstrating that market-rule priors and expressive neural networks should be jointly used for accurate and computationally sustainable forecasting in industrial energy trading applications. The implementation is publicly available at https://runyao-yu.github.io/MRINN/.

---

### Language-Conditioned Visual Grounding with CLIP Multilingual

Authors: J. de Curtò, Mauro Liz, I. de Zarzà
Published: 2026-05-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2605.09060v1

Multilingual vision-language models exhibit systematic performance gaps across languages, but the mechanism remains ambiguous: cross-language divergence could arise from the visual encoder, the text branch, or their interaction. We resolve this ambiguity through a dense multilingual CLIP probe in which the visual encoder is held identical across thirteen typologically diverse languages and only the XLM-RoBERTa text branch varies. We evaluate two CLIP architectures spanning a 7x visual-encoder scale gap (XLM-R base + ViT-B/32, ~87M visual parameters; XLM-R large + ViT-H/14, ~632M) on 11 concepts and 210 images, and quantify cross-language agreement via cluster-mask IoU, top-percentile IoU, and Spearman rank correlation against an English reference (n=2,310 paired observations per language). Three findings emerge. First, low-resource languages (Arabic, Basque, Luxembourgish) incur a structural penalty at both backbone scales (Wilcoxon HR>LR p<10^-300; cluster-mask IoU gap +0.114 at base, +0.143 at large), isolating the deficit to the text branch. Second, scaling the encoder 7x widens the gap for structural failure cases (Basque Δ=-0.056, Luxembourgish Δ=-0.076) while improving Arabic (Δ=+0.033), separating corpus-coverage from tokeniser-fertility failures. Third, peak similarity is preserved across languages (mean ratio 0.94 at large scale) while cluster-mask IoU drops sharply, identifying spatial misalignment, not signal collapse, as the dominant failure mode. At 3.4-3.9 Wh per 1,000 queries, dense-CLIP grounding is competitive with high-throughput inference budgets, positioning it as a practical substrate for energy-aware multilingual deployment.

---

### Nonlinear GENERIC-Embedded Neural Networks (N-GENNs): Learning GENERIC dynamics with non-quadratic dissipation potentials

Authors: Vojtěch Votruba, Zequn He, Weilun Qiu, Celia Reina, Michal Pavelka
Published: 2026-05-09
Categories: physics.comp-ph, cs.LG
arXiv: https://arxiv.org/abs/2605.09058v2

We introduce Nonlinear GENERIC-Embedded Neural Networks (N-GENNs), a deep learning framework for discovering evolution equations of systems governed by the nonlinear GENERIC formalism (General Equation for Non-Equilibrium Reversible-Irreversible Coupling). Such systems exhibit coupled conservative and dissipative dynamics, and can be described via the superposition of a Hamiltonian flow and a generalized gradient flow. In contrast to existing approaches, our formulation incorporates generalized gradient flows via convex dissipation potentials, enabling the identification of a broader class of thermodynamically consistent dynamics, including systems with non-quadratic dissipation potentials. Thermodynamic structure is strongly enforced by construction through suitable reparameterizations of both the reversible operator and the dissipation potential, ensuring exact compliance with the first and second laws of thermodynamics. We validate the proposed approach on three representative examples: a harmonic oscillator coupled to a heat bath, an idealized chemical motor, and a one-dimensional viscoplastic model of Perzyna type. These results demonstrate the method's ability to accurately infer thermodynamically consistent models from data for systems incorporating both conservative and nonlinear dissipative dynamics.

---

### Nonlinear GENERIC Informed Neural Networks (N-GINNs): learning GENERIC dynamics with non-quadratic dissipation potentials

Authors: Vojtěch Votruba, Zequn He, Weilun Qiu, Celia Reina, Michal Pavelka
Published: 2026-05-09
Categories: physics.comp-ph, cs.LG
arXiv: https://arxiv.org/abs/2605.09058v1

We introduce Nonlinear GENERIC Informed Neural Networks (N-GINNs), a deep learning framework for discovering evolution equations of systems governed by the nonlinear GENERIC formalism (General Equation for Non-Equilibrium Reversible-Irreversible Coupling). Such systems exhibit coupled conservative and dissipative dynamics, and can be described via the superposition of a Hamiltonian flow and a generalized gradient flow. In contrast to existing approaches, our formulation incorporates generalized gradient flows via convex dissipation potentials, enabling the identification of a broader class of thermodynamically consistent dynamics, including systems with non-quadratic dissipation potentials. Thermodynamic structure is strongly enforced by construction through suitable reparameterizations of both the bivector operator and the dissipation potential, ensuring exact compliance with the first and second laws of thermodynamics. We validate the proposed approach on three representative examples: a harmonic oscillator coupled to a heat bath, an idealized chemical motor, and a one-dimensional viscoplastic model of Perzyna type. These results demonstrate the method's ability to accurately infer thermodynamically consistent models from data for systems incorporating both conservative and nonlinear dissipative dynamics.

---

### Octopus Protocol: One-Shot Hardware Discovery and Control for AI Agents via Infrastructure-as-Prompts

Authors: Quilee Simeon, Justin M. Wei, Yile Fan
Published: 2026-05-09
Categories: cs.RO, cs.AI, cs.MA
arXiv: https://arxiv.org/abs/2605.09055v2

Bringing a previously unintegrated device under the control of an AI agent still requires device-specific engineering: driver selection, dependency resolution, interface design, and deployment, repeated per device and per platform. We present Octopus, a hardware onboarding framework in which a coding agent, rather than a shipped integration, is the runtime that produces the required infrastructure. Given shell access and a model API key, a single bootstrap command drives the agent through a five-stage pipeline that enumerates operating-system- visible hardware, infers device identity and capabilities, generates typed Model Context Protocol tools and the hardware-facing code behind them, and activates the result as a live endpoint. A persistent daemon then maintains the result, repairing defined classes of failure in the deployment it produced. Across four hosts spanning two processor architectures, three operating- system families, and two device-access paths, identical prose specifications produced working interfaces with no per-host edits and no hand-written integration code. Five consecutive runs on the reference host completed end to end on first attempt. We report both the resulting capability and a failure mode of unattended repair loops observed over eleven hours of continuous operation.

---

### Octopus Protocol: One-Shot Hardware Discovery and Control for AI Agents via Infrastructure-as-Prompts

Authors: Quilee Simeon, Justin M. Wei, Yile Fan
Published: 2026-05-09
Categories: cs.RO, cs.AI, cs.MA
arXiv: https://arxiv.org/abs/2605.09055v1

Recent agentic-robotics systems, from Code-asPolicies to modern vision-language-action (VLA) foundation models, presuppose that drivers, SDKs, or ROS-style primitives for the target hardware already exist. Writing those primitives is the dominant engineering cost of bringing up new hardware for agent control. We present Octopus Protocol, a system that collapses that cost to a single shell command. Given only raw OS access and a language-model API key, a coding agent executes a five-stage pipeline--PROBE, IDENTIFY, INTERFACE, SERVE, DEPLOY--to discover connected devices, infer their capabilities, generate a Model Context Protocol (MCP) server with typed tools, and deploy it as a live HTTP endpoint. A persistent daemon then monitors the system, heals broken code, and perceives physical state through the camera tools it generated for itself. Two architectural principles make this work: protocols are prompts, not code, and the coding agent is the runtime. We validate the system on three heterogeneous platforms (PC/WSL, Apple Silicon macOS, Raspberry Pi 4) and on a commercial 6-DOF robotic arm with USB camera feedback. One command onboards the hardware in ~10-15 minutes and exposes up to 30 MCP tools; an MCP-compliant client then performs closed-loop visual-motor control through tools no human wrote.

---