探索全球前沿学术脉络

01.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.15912

On-Policy Distillation with Curriculum Turn-level Guidance for Multi-turn Agents

作者:

Gengsheng Li↗Mao Zheng↗Mingyang Song↗Ruiqi Liu↗Tianyu Yang↗Jie Sun↗Qiyong Zhong↗Haiyun Guo↗Junfeng Fang↗Dan Zhang↗Jinqiao Wang↗

arXiv:2606.15912v1 Announce Type: cross Abstract: Multi-turn agents that plan, invoke tools, and interact with environments offer a promising paradigm for solving complex tasks, yet their capabilities typically rely on very large models whose inference cost is prohibitive in practice.On-Policy Distillation (OPD) is a natural recipe for transferring such capabilities to smaller students, but we find that it suffers a characteristic failure mode in this setting: small student errors compound across turns and push the trajectory out of the teacher's familiar state distribution, so the teacher's supervision becomes least reliable precisely where the student needs it most.We propose Guided On-Policy Distillation (Guided-OPD), a simple yet effective algorithm that mixes teacher- and student-generated turns within each rollout and schedules the teacher's intervention probability along a curriculum that decays to zero.Strong guidance keeps early trajectories close to the teacher distribution and is then gradually withdrawn to recover the purely on-policy regime used at inference.On ALFWorld, ScienceWorld, and WebShop, distilling Qwen3 students from a Qwen3-30B-A3B teacher, Guided-OPD improves Score by 21.1\% and Success Rate by 25.5\% over vanilla OPD on average, with larger gains on smaller students.

阅读与讨论 → 访问原文 →

02.

arXiv (CS.AI) 2026-06-24 DOI: arXiv:2606.24669

LaGO: Latent Action Guidance for Online Reinforcement Learning

作者:

Kuan-Yen Liu↗Ren-Jyun Huang↗Ti-Rong Wu↗

arXiv:2606.24669v1 Announce Type: new Abstract: Large language models (LLMs) have shown strong potential for planning and sequential decision-making, but prior work often relies on using them as direct controllers, which requires precise action generation and can be unreliable in practice. This paper proposes Latent Action Guidance for Online Reinforcement Learning (LaGO), a framework that uses a pretrained LLM as a latent action prior to softly guide online policy optimization, rather than treating the LLM as an explicit planner or controller. Experiments on both a discrete-control benchmark, CLEVR-Robot, and a continuous-control benchmark, Meta-World, demonstrate that LaGO consistently improves both reward and success rate over Vanilla PPO. In particular, LaGO increases the average success rate from 15.1% to 27.2% on CLEVR-Robot and from 2.7% to 15.2% on Meta-World. Our analysis further shows that stronger pretrained LLMs provide more effective guidance, suggesting that LLM knowledge can improve planning and online decision-making.

阅读与讨论 → 访问原文 →

03.

medRxiv (Medicine) 2026-06-16 DOI: HASH:c8e588944deee2a4b125eef7ffc98784

Recurrence After Hepatic Hydatid Cyst Surgery: Scolicidal Agent Application Technique and the Effect of Cystopiliary Fistula

作者:

Erdogan↗Kol Özbay↗Özbay↗M. A↗

Objective: This study aimed to evaluate long-term outcomes in patients who underwent surgical treatment for hepatic hydatid cyst (HCC) disease and, in particular, to investigate the effect of scolicidal agent (SA) application method and the presence of cystobiliary fistula (CBF) on the development of recurrence. Materials and Methods: This single-center, retrospective study included 197 patients who underwent surgical treatment for HCC disease. Hypertonic saline was used as SA in all patients and was classified as intracystic or pericystic application according to the application method. The presence of CBF was evaluated according to intraoperative and postoperative findings. Patients were followed for 86 months, and the development of recurrence was identified by radiological methods. Comparisons were made between the groups with and without recurrence in terms of SA application method and the presence of CBF. Results: The median age of the patients was 38 years, and the median follow-up period was 86 months. SA application was performed into the cyst in 51.3% of the patients and around the cyst in 48.7%. The presence of CBF was detected in 49.7% of the patients. No statistically significant difference was found between the recurrent and non-recurrent groups in terms of SA application method (p = 0.344). Similarly, no significant relationship was found between the presence of CBF and the development of recurrence (p = 0.721). Conclusion: This study showed that the SA application method and the presence of CBF are not determinants of recurrence in HCC disease. It is thought that recurrence rates can be kept low with appropriate surgical technique and effective biliary tract management.

阅读与讨论 → 访问原文 →

04.

arXiv (CS.AI) 2026-06-24 DOI: arXiv:2605.24050

More Skills, Worse Agents? Skill Shadowing Degrades Performance When Expanding Skill Libraries

作者:

Hongwen Song↗Song Wei↗

arXiv:2605.24050v2 Announce Type: replace-cross Abstract: Skill libraries allow LLM agents to load task-specific instructions on demand, letting non-expert users solve domain-specific tasks through natural language without knowing which skills exist or how they work. However, performance degrades as libraries grow – by up to 21\% when scaling from a small set of helpful skills to a 202-skill library. In this work, we formulate this performance degradation as the pass rate drop between loading a library of known-helpful skills and the full library. Moreover, we propose to decompose the pass rate drop by conditioning on the skill(s) invocation – which skills the agent selects during a trajectory – into two effects: skill shadowing, where the agent selects wrong skills more often as the library expands, and context overhead, where the enlarged context degrades execution even when selection is correct. We derive upper bounds on both effects to characterize their magnitudes of impacts to the pass rate drop. Our empirical estimates of the effects and their upper bounds both show that the skill shadowing effect grows with library size and significantly contributes to the performance degradation, whereas the context overhead effect remains small and indistinguishable from zero. This observed asymmetry establishes that the skill selection failure, not the enlarged context, is the primary bottleneck when expanding the skill libraries.

阅读与讨论 → 访问原文 →

05.

arXiv (math.PR) 2026-06-11 DOI: arXiv:2606.12143

Continuous stochastic flows driven by white noise and their duals

作者:

Yaolin Yu↗

arXiv:2606.12143v1 Announce Type: new Abstract: We study a class of continuous stochastic flows driven by a space-time white noise and characterize their dual flows by explicit stochastic differential equations. A key ingredient of the proof is the convergence of solutions under coefficient approximations. As an application, we derive the dual flows in two illustrative examples, the squared Bessel flow and the Jacobi flow. We also introduce a new model of polynomially self-repelling (PSR) flow and show that it enjoys a self-duality property.

阅读与讨论 → 访问原文 →

06.

arXiv (CS.AI) 2026-06-11 DOI: arXiv:2601.11670

CoVar: Confidence-Variance-Guided Pseudo-Label Selection for Semi-Supervised Learning

作者:

Jinshi Liu↗Lei He↗Pan Liu↗

arXiv:2601.11670v3 Announce Type: replace-cross Abstract: Pseudo-label selection in semi-supervised learning is commonly driven by maximum-confidence thresholds, yet confidence alone can be unreliable under model overconfidence and class imbalance. We propose CoVar, a confidence–variance framework that assesses pseudo-label reliability by jointly modeling Maximum Confidence (MC) and Residual-Class Variance (RCV). Starting from entropy minimization, we derive a second-order cross-entropy approximation showing that low-loss pseudo-labels are favored when MC is high and RCV is low, with a confidence-dependent penalty that becomes stronger for near-certain predictions. Based on this criterion, CoVar embeds predictions into a two-dimensional confidence–variance space and uses SVD-based spectral relaxation to separate reliable and unreliable predictions without hand-tuned confidence thresholds. Cluster-wise Gaussian weighting then converts this separation into per-sample training weights. The resulting weights can be integrated into existing semi-supervised segmentation and classification pipelines during training and introduce no inference-time overhead. Experiments on PASCAL VOC 2012, Cityscapes, CIFAR-10, CIFAR-100, SVHN, and STL-10 show clear gains on VOC and Cityscapes under matched backbones, as well as competitive or improved error rates on standard classification benchmarks. These results indicate that residual-class dispersion provides a useful signal complementary to confidence for robust pseudo-label selection.

阅读与讨论 → 访问原文 →

07.

arXiv (quant-ph) 2026-06-25 DOI: arXiv:2606.25795

Benchmarking Dark Matter Search using a Parity-Check Protocol with Machine-Learning Optimized Pulses

作者:

Yu-Han Chang↗Ilya Moskalenko↗Marko Kuzmanovi\'c↗Ognjen Stanisavljevi\'c↗Isak Bj\"orkman↗David D\'iez-Ib\'a\~nez↗Yikun Gu↗Akash V. Dixit↗Igor G. Irastorza↗Gheorghe Sorin Paraoanu↗

arXiv:2606.25795v1 Announce Type: new Abstract: We report on an improved microwave detection protocol for dark matter candidates such as the axion and the dark photon. We employ a superconducting transmon qubit dispersively coupled to a double-cavity system, enabling quantum non-demolition measurements of the photon occupation in a relatively short-lived storage cavity. To reduce the experimental cycle time and enhance sensitivity for axion and dark-photon searches, we operate this detector in a regime of increased qubit-cavity coupling, resulting in Stark shifts of 4.6 MHz. In this regime, conventional control pulses suffer from strong frequency-detuning sensitivity and photon-number-dependent errors. We address this limitation by implementing frequency-detuning-robust $\pi/2$ pulses (obtained by machine-learning optimization) that preserve high-fidelity qubit control over a bandwidth of approximately 20 MHz. We experimentally validate this protocol and demonstrate single-photon detection performance comparable to previous implementations, despite significantly reduced qubit coherence times and storage-cavity lifetimes. Using parity-based measurement sequences combined with a Hidden Markov Model (HMM) analysis, we achieve background rates on the order of $\mathcal{O}(20)$ Hz. In the absence of a magnetic field, we derive exclusion limits on the dark photon model for dark matter, reaching a sensitivity to the kinetic mixing angle of $\epsilon_{95\%} \sim 1\times10^{-14}$ at 5.051 GHz. These results establish machine-learning robust control as a key enabler for faster, more scalable microwave quantum sensors for dark-matter searches.

阅读与讨论 → 访问原文 →

08.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.16989

Stable Menus of Public Goods: AI-Enabled Progress

作者:

Sara Fish↗

arXiv:2606.16989v1 Announce Type: cross Abstract: Using an open problem from the EC 2025 paper "Stable Menus of Public Goods" as a testbed, we conduct experiments to understand the effectiveness of different AI-for-EconCS research workflows. Specifically, we study three questions: Does providing human intuition in the prompt help? Does automated multi-turn interaction help? And, does an LLM outperform a first-year PhD student? Regarding the first two questions, we provide evidence for the following workflow suggestions: (1) prompting with human intuition can encourage the LLM to have better "taste", (2) multi-turn workflows help when the pipeline encourages "ambitious" steps. Regarding the third question, using an unpublished manuscript written by the paper's senior authors prior to collaborating with the first-year PhD student, we compare the effectiveness of the LLM with that of the first-year PhD student, and find that the LLM is slightly less effective.

阅读与讨论 → 访问原文 →

09.

arXiv (CS.LG) 2026-06-25 DOI: arXiv:2606.25265

Variational Inference via Entropic Transport Descent

作者:

Vincent Pacelli↗Akash Ratheesh↗Evangelos Theodorou↗

arXiv:2606.25265v1 Announce Type: new Abstract: Particle-based variational inference (ParVI) methods approximate an intractable target distribution by evolving an ensemble of interacting samples. Existing approaches rely predominantly on kernel-based repulsion (e.g., SVGD), which suffers from variance collapse in high dimensions and mode collapse on multimodal targets – pathologies caused by the absence of global transport structure. We introduce entropic transport descent (ETD), a ParVI family that frames each particle update as an entropy-regularized optimal transport problem. Derived from the JKO proximal scheme by lifting to the space of couplings and relaxing via the KL chain rule, each ETD iteration reduces to a Sinkhorn computation. The resulting transport plan provides global coordination, guiding each particle to nearby high-density proposals and naturally preserving multimodal structure. ETD can operate entirely score-free, requiring only pointwise evaluations of the unnormalized target density. Experiments on variance-collapse diagnostics, Bayesian logistic regression, neural networks, and molecular Boltzmann distributions show that ETD matches or outperforms SVGD, AGF-SVGD, and SGLD, with the largest gains in high-dimensional and multimodal settings.

阅读与讨论 → 访问原文 →

10.

arXiv (CS.LG) 2026-06-12 DOI: arXiv:2508.04888

Retrieval-Augmented Foundation Models for Water Level Prediction in the Everglades

作者:

Rahuul Rangaraj↗Jimeng Shi↗Rajendra Paudel↗Giri Narasimhan↗Yanzhao Wu↗

arXiv:2508.04888v2 Announce Type: replace Abstract: Accurate water level forecasting in the Everglades is essential for flood mitigation, drought management, water resource planning, and biodiversity conservation. While recent time-series foundation models have shown strong performance on generic tasks (represented in their pre-training), their effectiveness in domain-specific applications remains insufficiently understood. In this work, we curate a domain-specific dataset for water-level forecasting in the Everglades and observe that the performance of current state-of-the-art models remains limited. To address this gap, we leverage a retrieval-augmented mechanism that retrieves analogous multivariate hydrological episodes from an external archive of historical observations to enrich the input context of those pre-trained models. We study two retrieval strategies, statistical similarity-based retrieval and mutual information-based retrieval, and analyze how incorporating retrieved historical contexts affects predictive performance. Extensive experiments show that retrieval augmentation consistently improves long-horizon water level forecasts and yields disproportionately larger gains during extreme events, which is particularly critical for environmental decision-making. Our study provides empirical evidence that analog-based retrieval can benefit pretrained time-series foundation models in environmental science, offering practical insights into their strengths, limitations, and failure modes when applied to hydrological forecasting in the Everglades. Although evaluated in the Everglades, the proposed framework is general and can be applied to other hydrological systems given time series data. The code and data have been made publicly available at https://github.com/rahuul2992000/WaterRAF.

阅读与讨论 → 访问原文 →

11.

arXiv (CS.AI) 2026-06-25 DOI: arXiv:2606.25401

Lightweight PCGAE-Net: Parallel CrossGate Attention and Bottleneck AutoEncoder for Efficient 5G Channel Prediction

作者:

Uma Kishore Godavarti↗K. Giridhar↗Vanani Prince Dharmendrabhai↗Anchit Panday↗Madhan Raj Kanagarathinam↗

arXiv:2606.25401v1 Announce Type: cross Abstract: Accurate channel state information (CSI) prediction is essential for proactive beamforming and resource management in 5G massive MIMO systems, yet the deployment of high-accuracy transformer-based predictors on base-station hardware remains challenging because the most capable models carry upwards of 30\,M parameters. This paper introduces Lightweight PCGAE-Net, which addresses the efficiency problem not by post-hoc compression but by correcting two architectural flaws in the current state of the art. The first is a sequential attention ordering bias: in CS3T-UNet, group-wise temporal attention (GTA) always operates on features that have already been transformed by cross-shaped spatial attention (CSA), distorting what temporal information GTA can capture. We remove this dependency by routing both attention modules to the same layer-normalized input and combining their independent outputs through a learned per-channel sigmoid CrossGate. The second flaw is an uncompressed bottleneck: applying full self-attention at the deepest encoder stage, where channel depth reaches $4C$, is quadratically expensive and carries redundant features. A Bottleneck AutoEncoder (BAE) with $1\times1$ convolutions halves this depth and uses an auxiliary reconstruction loss to prevent information collapse. Wrapping these components inside a shallower encoder-decoder with frequency-domain dimensionality reduction ($N_f\!=\!32$, $C\!=\!48$) produces a model with just 8.54\,M parameters – 58\% fewer than the CS3T-UNet baseline – that outperforms it by up to 3.26\,dB at 5\,km/h and 6.0\,dB at 9\,km/h in single-step prediction on QuaDriGa dataset.

阅读与讨论 → 访问原文 →

12.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2605.01973

Learn-to-learn on Arbitrary Textual Conditioning: A Hypernetwork-Driven Meta-Gated LLM

作者:

Luo Ji↗Qi Qin↗Ningyuan Xi↗Teng Chen↗Qingqing Gu↗Hongyan Li↗

Conventional LLMs may suffer from corpus heterogeneity and subtle condition changes. While finetuning can create the catastrophe forgetting issue, application of meta-learning on LLMs is also limited due to its complexity and scalability. In this paper, we activate the meta-signal of $\beta$ within the SwiGLU blocks, resulting in a meta-gating mechanism that adaptively adjusts the nonlinearity of FFN. A hypernetwork is employed which dynamically produces $\beta$ on textual conditions, providing meta-controllability on LLMs. By testing on different condition types such as task, domain, persona, and style, our method outperforms finetuning and meta-learning baselines, and can generalize reasonably on unseen tasks, condition types, or instructions. Our code can be found in https://github.com/AaronJi/MeGan.

阅读与讨论 → 访问原文 →

13.

arXiv (CS.LG) 2026-06-25 DOI: arXiv:2606.25488

Distill on a Diet: Efficient Knowledge Distillation via Learnable Data Pruning

作者:

Yifan Wu↗Yiqi Wang↗Xichen Ye↗Wenjing Yan↗Xiaoqiang Li↗Cheng Jin↗Xiangyu Yue↗Weizhong Zhang↗

arXiv:2606.25488v1 Announce Type: new Abstract: Knowledge Distillation (KD) is widely used to obtain compact models for efficient inference in resource-constrained environments. Yet the computational overhead of the distillation process itself is often overlooked, raising the question of whether a better student model can be obtained with less data and less compute via data pruning. However, existing data pruning methods are not designed for KD: some introduce substantial overhead, such as obtaining training dynamics through retraining, while others rely on heuristic selection rules that fail to capture what KD actually requires, often resulting in suboptimal subsets. To address these issues, we propose IF-Beta, an efficient data pruning framework that combines influence functions with a learnable sampling policy. Empirically, we first demonstrate that influence functions can serve as an effective and efficient estimator of sample impact in KD settings, where only a pretrained teacher is available. Building on this, our sampling policy is specifically parameterized by a Beta distribution, whose highly flexible two-parameter family allows the policy to adapt to diverse pruning regimes rather than being tied to fixed heuristic forms. Next, we formulate KD pruning as optimizing this policy through a bilevel objective, where the inner loop operates in the teacher feature space with a KD-aligned objective, enabling fast proxy training, while the outer loop updates the policy parameters to maximize distillation performance. This design ensures that IF-Beta is both computationally efficient and inherently aligned with the goals of KD. Extensive experiments on CIFAR-10/100 and ImageNet show that IF-Beta consistently outperforms other baselines across a wide range of pruning ratios. Remarkably, IF-Beta enables students trained on less data and less compute to surpass the performance of students distilled on the full dataset.

阅读与讨论 → 访问原文 →

14.

arXiv (quant-ph) 2026-06-25 DOI: arXiv:2606.25920

Finite-Shot Sensitivity for Moment Estimation in Quantum Metrology

作者:

Shaowei Du↗Shuheng Liu↗Weidong Li↗Luca Pezz\`e↗Augusto Smerzi↗Qiongyi He↗

arXiv:2606.25920v1 Announce Type: new Abstract: The quantum Cramér-Rao bound can be saturated only asymptotically and does not specify how many measurements are needed for a concrete estimator to approach it. We develop a finite-measurement theory for method-of-moments estimation, where the parameter is inferred from the sample mean of a calibrating observable rather than from the full likelihood. For general quantum statistical models, the expansion is written in terms of the calibration curve and the central moments of the measured observable. Nonlinear calibration curves make the usual moment estimator biased at finite measurement number; we construct a bias-corrected estimator with bias $O(\nu^{-3})$. This gives sensitivity corrections beyond the leading error-propagation term of the chosen moment protocol. We identify a general density-matrix condition under which the full $1/\nu^2$ correction vanishes. In unitary examples, the leading residual correction appears at order $1/\nu^3$, is governed by calibration curvature, and can be reduced or cancelled by higher-rank components of the same measured observable. The resulting thresholds quantify how many measurements are needed before the asymptotic sensitivity of a moment-estimation protocol is operationally visible.

阅读与讨论 → 访问原文 →

15.

arXiv (CS.CL) 2026-06-15 DOI: arXiv:2606.13712

Multimodal Speaker Identification in Classroom Environments

作者:

Michael L. Chrzan↗Meghavarshini Krishnaswamy↗Robert Gibboni↗Katie Wetstone↗Wei Ai↗Jing Liu↗

Automated analysis of K-12 classroom dynamics faces challenges due to background noise and variable child speech, often confounding acoustic-only models. This study evaluates a multimodal speaker identification framework anchoring acoustic embeddings with LLM-derived semantic context. Using a subset of the EDSI dataset (8 math classrooms, N = 2,801 utterances), we found an acoustic baseline (ECAPA-TDNN) achieved only 39.0% accuracy. By integrating transcript-based "contextual anchoring" into a gradient boosting classifier, our multimodal approach raised student identification to 50.3%. Performance also improved for utterances over 5 seconds, reaching 76.9% accuracy (vs. 64.9% baseline) with a 90.9% Top-3 accuracy. Additionally, the model distinguished teacher vs. student roles with 99.3% accuracy. This approach advances the feasibility of automated feedback systems capable of considering individual student participation, a crucial step for supporting equitable instruction at scale.

阅读与讨论 → 访问原文 →

16.

arXiv (quant-ph) 2026-06-16 DOI: arXiv:2512.16242

Hardy-type self-testing and exposedness of tripartite GHZ correlations

作者:

Smritikana Patra↗Soumyajit Pal↗Ranendu Adhikary↗

arXiv:2512.16242v2 Announce Type: replace Abstract: Nonlocality can be witnessed either through Bell-inequality violations or through logical contradictions such as Hardy's paradox. In the bipartite two input two outcome scenario, these two routes have distinct geometric behavior: CHSH-maximal correlations are exposed points of the quantum set, whereas known Hardy-type self-testing correlations on the no-signaling boundary are non-exposed. Here we show that this bipartite intuition fails in the tripartite two input two outcome scenario. We study the tripartite instance of a multipartite Hardy-type paradox and prove that the correlation attaining the maximal Hardy success probability self-tests the Greenberger–Horne–Zeilinger state and the associated measurements. Although this correlation lies on the no-signaling boundary, we show that it is an extremal and exposed point of the quantum correlation set. Moreover, it coincides with the correlation attaining the maximal violation of the Mermin inequality. Thus, in the tripartite GHZ scenario, the logical-paradox and Bell-inequality routes to nonlocality select the same exposed quantum boundary point. We also establish a robust version of the self-test, showing that small deviations from the ideal Hardy constraints imply quantitative closeness to the target state and measurements. Our results reveal a qualitative geometric difference between bipartite and tripartite Hardy-type nonlocality and suggest a broader investigation of exposedness for multipartite Hardy correlations in the multiparty setting.

阅读与讨论 → 访问原文 →

17.

arXiv (CS.CL) 2026-06-17 DOI: arXiv:2606.17861

GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine?

作者:

Tongxu Luo↗Rongsheng Wang↗Jiaxi Bi↗Chenming Xu↗Zhengyang Tang↗Jianlong Chen↗Juhao Liang↗Ke Ji↗Shuqi Guo↗Yuhao Du↗Fan Bu↗Wenyu Du↗…

Game generation is an emerging application of coding agents, requiring models to transform natural-language specifications into playable interactive systems. Unlike traditional coding tasks, game generation takes place within a game engine, where scripts, scenes, assets, rendering, and runtime interactions must jointly produce coherent gameplay. We formalize end-to-end game generation as the problem of producing a complete game artifact that realizes a specification through observable player-game interaction in a target environment. We argue that evaluating this setting requires three desiderata: Engine Grounding, Artifact Completeness, and Interactive Verification. We propose an interaction-grounded evaluation framework that assesses executable gameplay through replayed demonstrations and rubric-guided multimodal judging. We instantiate this framework as GameCraft-Bench, a benchmark comprising 140 Godot tasks across 15 game families. Evaluations of frontier coding agents show that end-to-end game generation remains highly challenging: the strongest agent achieves only 41.46%, and most agents score below 40%. Further analysis reveals that while agents often implement recognizable mechanics, they struggle to deliver complete games with sufficient content, functional visual feedback, and coherent presentation. See https://tongxuluo.github.io/gamecraft-bench-website for demos, code, and data.

阅读与讨论 → 访问原文 →

18.

arXiv (CS.CL) 2026-06-25 DOI: arXiv:2606.25556

BiPACE: Bisimulation-Guided Policy Optimization with Action Counterfactual Estimation for LLM Agents

作者:

Hanyang Wang↗Weijieying Ren↗Yuxiang Zhang↗Ding Cao↗Zhizhao Zeng↗Ke Zeng↗Tianxiang Zhao↗

Stepwise group-based RL is an attractive way to train long-horizon LLM agents without a learned critic: it reuses multiple sampled rollouts to estimate local advantages. Its weakness is less visible but more fundamental: every group-relative estimator assumes that the steps it compares are equivalent for credit assignment. We show that current agentic variants violate this assumption through a state-action credit mismatch. The observation-hash partition is overly fine on the state side, creating singleton groups with zero step-level signal, while a single within-group mean is too coarse on the action side, mixing state-value estimation with action-specific credit. We introduce BiPACE (Bisimulation-Guided Policy Optimization with Action Counterfactual Estimation), a drop-in advantage estimator that fixes both sides without adding a critic, auxiliary loss, or extra rollouts. BiGPO clusters steps by cosine distance in the actor's own hidden-state geometry, an empirical policy-induced proxy for bisimulation that substantially lowers the singleton rate left by observation hashing. PACE then recenters returns within each behavioral cluster using action-conditioned peer baselines; its Q-style instance estimates a local Q(s,a)-V(s) nonparametrically. On ALFWorld/Qwen2.5-7B, BiPACE_Q raises overall validation success from GiGPO's 90.8 to $97.1\pm0.9$ over three seeds, and crosses the 95% threshold on every seed, which GiGPO never does within the same budget. On Qwen2.5-1.5B it reaches $93.5\pm1.2$ versus GiGPO's 86.7, and on WebShop and TextCraft it improves over GRPO and GiGPO at both model scales. The measured BiPACE-specific overhead is 11.3% of a single training-step wall time. Yet it changes the estimator's comparison unit from surface identity to approximate behavioral equivalence plus action-side counterfactuals. The code is available at https://github.com/TianxiangZhao/BiPACE.

阅读与讨论 → 访问原文 →

19.

arXiv (CS.AI) 2026-06-15 DOI: arXiv:2606.13832

Safety-Contract Graph Multi-Agent Reinforcement Learning for Autonomous Network Security Response

作者:

Jose Luis Lima de Jesus Silva↗

arXiv:2606.13832v1 Announce Type: cross Abstract: Autonomous network-security response systems promise to reduce Security Operations Centre (SOC) reaction latency, but reward-only multi-agent reinforcement learning (MARL) can improve security reward while remaining non-deployable. We present a safety-contract graph MARL framework and instantiate it as ACD$^3$-GAT (Adaptive Constrained Counterfactual Decisioning with a Graph Attention Network encoder), an architecture that separates simulator observations from reusable operational budgets, constrained optimization, graph state encoding, and counterfactual action screening. We evaluate the method in CAGE Challenge 4, where agents operate under budgets for Mean Time to Recover (MTTR), false-positive response, and firewall change-management disruption. Across the benchmark, every unconstrained method violates the SOC downtime budget in 100% of evaluated episodes, with mean downtime proxy costs of 311-430 against a budget of 50. This complements prior CAGE Challenge 4 findings by showing that reward-only learning lacks operational discipline. Constrained MAPPO-GAT (C-MAPPO-GAT) isolates Lagrangian operational-cost control and budget-aware screening, while ACD$^3$-GAT adds budget context, CVaR tail-risk estimation, opponent-belief state, and Graph Counterfactual Risk Propagation (G-CRP). The replicated comparison includes three 200-episode seeds for IPPO, MAPPO-GAT, C-MAPPO-GAT, and ACD$^3$-GAT. C-MAPPO-GAT reduces downtime violation from 100% to 0.3% and mean downtime cost from 355.4 to 15.5 relative to MAPPO-GAT. ACD$^3$-GAT reduces mean downtime cost to 48.2 with a 13.8% violation rate, placing it on the safety-contract frontier rather than at the most conservative compliance point. Topology-seed and coupled adaptive Red-process stress tests preserve this contrast and show lower worst adaptive degradation for safety-constrained policies than reward-only MAPPO-GAT.

阅读与讨论 → 访问原文 →

20.

arXiv (CS.AI) 2026-06-12 DOI: arXiv:2606.13020

SciR: A Controllable Benchmark for Scientific Reasoning in LLMs

作者:

Pierre Beckmann↗Marco Valentino↗Andre Freitas↗

arXiv:2606.13020v1 Announce Type: new Abstract: Three paradigmatic forms of inference recur across scientific reasoning: deduction, induction, and causal abduction. Reliably evaluating LLMs on these in scientific settings is currently out of reach: scientific benchmarks built on human annotations are costly and lack mechanistic ground truth, while synthetic logical-reasoning benchmarks do not resemble real scientific documents. We introduce SciR, a benchmark that combines multi-paradigm reasoning with controllable scientific rendering, anchored on three paradigmatic scientific problems. Tasks are generated from formal objects (deduction tree, inductive rule hypothesis, causal graph) to guarantee verifiable answers, then rendered into multi-document scientific discourse via per-track domain-tuned genres. The construction lets us independently vary two difficulty axes: how hard it is to extract the key information needed for inference, and how hard the principled inference itself is. We test six models. Both axes hurt every model, and their effects compound. The rendering even hurts neurosymbolic pipelines, which hand inference to a verified solver. The two axes yield a per-model extraction-vs-inference profile: for instance, reasoning models like deepseek-r1 mostly surpass non-reasoning instruct models on the inference axis. To our knowledge, SciR is the first multi-paradigm scientific-reasoning benchmark with parametric control on both extraction and inference difficulty.

阅读与讨论 → 访问原文 →

21.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.15206

AI Contagion in Social Networks

作者:

Olivier Bos↗Stefano Bosi↗

arXiv:2606.15206v1 Announce Type: cross Abstract: We study how artificial intelligence (AI) interacts with social communication networks to shape the stability of collective knowledge. Agents exchange information through a network while receiving AI-generated content, and AI systems retrain on the aggregate social information they influence. This interaction generates two feedback forces: an AI contagion channel, through which distortions diffuse across the network, and an AI social distortion multiplier, through which retraining amplifies past errors. Despite the high dimensionality of the environment, we show that the long-run behavior of the system admits a two-dimensional representation whose spectral radius determines whether AI-mediated information systems are dynamically stable or unstable. We characterize a sharp regulatory frontier identifying the minimum filtering required for stability and show how network topology shapes systemic informational risk.

阅读与讨论 → 访问原文 →

22.

arXiv (CS.CV) 2026-06-19 DOI: arXiv:2606.20545

Current World Models Lack a Persistent State Core

作者:

Jinpeng Lu↗Dexu Zhu↗Haoyuan Shi↗Linghan Cai↗Guo Tang↗Yinda Chen↗Jie Cao↗Duyu Tang↗Yi Zhang↗Yong Dai↗Xiaozhu Ju↗

World models are increasingly regarded as a decisive step toward artificial general intelligence, yet modeling the physical world demands more than rendering convincing frames on demand: it requires an internal world state that keeps evolving over time, decoupled from observation, so that objects endure and events run to their conclusions whether or not a camera is watching, much as the moon holds to its orbit when no one is looking. This requirement is a blind spot of existing benchmarks, which reward surface properties such as fidelity, motion, and camera controllability while never asking whether a generated world keeps evolving once it is unobserved. We introduce WRBench, the first systematic diagnostic benchmark that treats camera motion as an intervention on observability and resolves evaluation into a human-calibrated chain that asks whether the camera executes the requested interaction, whether the scene stays continuous and identifiable while in view, and whether a returning target remains consistent with the event that was set in motion. Across 9{,}600 videos from 23 models spanning four control paradigms, one finding proves stubborn: current systems maintain the observed world as a tracking shot, resuming a returning target in the state at which it was abandoned rather than advancing the event while it went unseen. Because this failure recurs across control paradigms, model families, and increments of scale, robust world-state evolution does not follow from cleaner imagery, tighter control, richer geometric priors, or sheer parameter count We therefore argue that the stability of the physical state kernel and the consistency of worldlines under viewpoint intervention should become first-class objectives of world-model design, so that a world model captures how the world will unfold rather than how the next frame appears.

阅读与讨论 → 访问原文 →

23.

arXiv (quant-ph) 2026-06-17 DOI: arXiv:2604.20912

Quantum-HPC Software Stacks and the openQSE Reference Architecture: A Survey

作者:

Amir Shehata↗Brian Austin↗Tom Beck↗Lukas Burgholzer↗Alex Chernoguzov↗Spencer Churchill↗Andrea Delgado↗Yasuko Eckert↗Jeffery Heckey↗Kevin Kissell↗Katherine Klymko↗Josh Moles↗…

arXiv:2604.20912v2 Announce Type: replace Abstract: Quantum resources are increasingly integrated into high-performance computing (HPC) and cloud environments, but quantum high-performance computing (QHPC) software stacks remain isolated, often proprietary, full-stack solutions lacking common interfaces across runtime, resource management, orchestration, and execution layers. This paper analyzes nine production QHPC stacks and identifies common design patterns and emerging requirements, covering deployment models, application interaction patterns, SDK support, and readiness for fault-tolerant operation. The survey exposes consistent needs in runtime abstraction, resource management, interconnect semantics, and observability. Based on these findings, we propose the open quantum-HPC software ecosystem ( openQSE) reference architecture as a first step toward unifying the state-of-the-practice. openQSE defines a set of layer boundaries that allow different implementations to interoperate while preserving deployment flexibility, and is structured to support both current noisy intermediate-scale quantum (NISQ) workloads and future fault-tolerant quantum computing (FTQC) systems without changes to upper-layer application interfaces.

阅读与讨论 → 访问原文 →

24.

arXiv (CS.CV) 2026-06-16 DOI: arXiv:2508.10523

Reasoning in Computer Vision: Taxonomy, Models, Tasks, and Methodologies

作者:

Ayushman Sarkar↗Zhenyu Yu↗Mohd Yamani Idna Idris↗

Visual reasoning matters for many computer vision tasks that go beyond surface-level object detection and classification. Despite progress in relational, symbolic, temporal, causal, and commonsense reasoning, existing surveys typically cover only one part of the problem, such as visual question answering, scene-graph generation, neuro-symbolic AI, or multimodal chain-of-thought, and rarely analyze reasoning types, methodologies, and evaluation protocols together. This survey addresses that gap. Following a structured literature review, we group visual reasoning into five major types (relational, symbolic, temporal, causal, and commonsense) and examine how each is implemented across methods that range from graph-based models, memory networks, attention mechanisms, and neuro-symbolic systems to reasoning with vision-language models (VLMs) and multimodal large language models (MLLMs), including visual chain-of-thought, visual programming, and tool-augmented and test-time reasoning. We then review evaluation protocols for functional correctness, structural consistency, and causal validity, and we analyze their limits in generalizability, reproducibility, faithfulness, and explanatory power. We also identify open challenges: scaling to complex scenes, integrating symbolic and neural paradigms more deeply, the shortage of comprehensive benchmarks, language-prior shortcuts and hallucination in foundation models, and reasoning under weak supervision. Finally, we set out a research agenda for vision systems and argue that connecting perception and reasoning is necessary for transparent, trustworthy, and cross-domain models, especially in high-stakes settings such as autonomous driving and medical diagnostics.

阅读与讨论 → 访问原文 →

25.

arXiv (CS.AI) 2026-06-24 DOI: arXiv:2505.24622

Random Rule Forest (RRF): Interpretable and Manageable Ensembles of LLM-Generated Questions for Predicting Success from Unstructured Data

作者:

Ben Griffin↗Aaron Ontoyin Yin↗Diego Vidaurre↗Ugur Koyluoglu↗Joseph Ternasky↗Fuat Alican↗Yigit Ihlamur↗

arXiv:2505.24622v3 Announce Type: replace Abstract: Many high-stakes screening tasks require predicting rare outcomes from unstructured text, where errors are costly and decisions must be auditable. We introduce Random Rule Forest (RRF), an interpretable ensemble that uses a large language model (LLM) not as an end-to-end predictor but as a generator of simple YES/NO questions. Each question acts as a weak learner, and their responses are combined by a plain unit-weight vote into an auditable ``green-flags'' scorecard: enough independent positive signals indicate a higher chance of success. We argue this deliberate simplicity is a robust default when positives are scarce and learned weights are hard to estimate. We evaluate RRF in two low-base-rate domains. On early-stage startup screening from founder profiles, RRF produces a transparent scorecard whose precision is several times the base rate (with light expert input raising it further) and, unlike direct prompting, its operating point can be controlled directly. On an established Phase~I clinical-trial benchmark, RRF outperforms published baselines on the threshold-independent metrics PR-AUC and ROC-AUC. Together these show that LLMs can serve as auditable feature generators for high-stakes text-based decisions, combining transparency with competitive predictive performance.

阅读与讨论 → 访问原文 →