Academic Intelligence · Curated Daily

探索全球前沿学术脉络

AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。

01.
Science (Express) 2026-05-07

TranscriptFormer: A generative cell atlas across 1.5 billion years of evolution | Science

作者: 未知作者

Single-cell transcriptomics is revolutionizing our understanding of cellular diversity, yet comparing transcriptional programs across the tree of life remains challenging. We developed TranscriptFormer, a family of generative foundation models trained on up to 112 million cells spanning 1.53 billion years of evolution across 12 species. We demonstrate state-of-the-art performance on cell type classification, even for species separated over 685 million years of evolution, and zero-shot disease state identification in human cells. Developmental trajectories, phylogenetic relationships and cellular hierarchies emerge naturally in TranscriptFormer’s representations without any explicit training on these annotations. This work establishes a powerful framework for quantitative single-cell analysis and comparative cellular biology, thus demonstrating that universal principles of cellular organization can be learned and predicted across the tree of life.

02.
arXiv (CS.AI) 2026-06-12

DailyReport: An Open-ended Benchmark for Evaluating Search Agents on Daily Search Tasks

arXiv:2606.12871v1 Announce Type: new Abstract: Search Agents (SAs) typically leverage large language models (LLMs) to support complex information-seeking tasks by autonomously exploring web sources and synthesizing information into comprehensive responses. For SAs evaluation, prior benchmarks mainly focus on specialized tasks that are unlikely to arise in real-world user scenarios. Moreover, their reliance on coarse task-level rubrics often limits evaluation interpretability. To bridge this gap, we introduce DailyReport, an open-ended benchmark to evaluate SA capabilities on daily search tasks. It contains 150 open-ended tasks with 3,546 associated rubrics, capturing widely discussed and timely information demands of real-world users. Each task is decomposed into subtasks and evaluated with cascade rubrics across disentangled dimensions. Through cascade performance attribution and user-centric aggregation, we derive highly interpretable scores for each dimension, along with a user preference score. Our results on 17 agentic systems show that current systems still fall short of users' expectations. To facilitate future research, our dataset and code are made publicly available at https://github.com/AGI-Eval-Official/DailyReport.

03.
medRxiv (Medicine) 2026-06-24

Topical fresh Taraxacum mongolicum wet dressing as an adjunct to ceftriaxone for localized skin and soft tissue infections: A single-center assessor-blinded randomized controlled trial

Background: Localized skin and soft tissue infections may need systemic antibacterials, but local inflammation can delay symptom recovery. We evaluated whether topical fresh Taraxacum mongolicum wet dressing added to ceftriaxone was associated with short-term benefit in selected clinically stable adults. Methods: In this single-center, assessor-blinded, three-arm randomized trial, 180 adults aged 18-74 years were randomized 1:1:1 to topical T. mongolicum plus intravenous ceftriaxone, topical T. mongolicum alone, or ceftriaxone alone for 7 days. The primary outcome was day-7 clinical response assessed by blinded independent assessors using prespecified global clinical improvement criteria. Analyses followed the intention-to-treat principle; sensitivity analyses assessed robustness. Results: Day-7 clinical response rates were 91.67% (55/60), 76.67% (46/60), and 68.33% (41/60) in the combined, T. mongolicum, and ceftriaxone groups, respectively (overall P = 0.006). Compared with ceftriaxone alone, combined therapy had a higher response rate (risk difference, 23.3 percentage points; 95% CI, 9.6 to 37.0; risk ratio, 1.34; 95% CI, 1.11 to 1.62). Sensitivity analyses were directionally consistent. Secondary outcomes and bacterial clearance favored the combined group. No serious adverse events were reported. Conclusions: In selected clinically stable adults with localized skin and soft tissue infections, adjunctive topical fresh T. mongolicum plus ceftriaxone was associated with improved short-term outcomes compared with ceftriaxone alone. Findings require cautious interpretation because this was a single-center, partially blinded trial without a placebo dressing control. The dressing should not replace antibiotics, drainage, or urgent care when indicated. Trial registration: International Traditional Medicine Clinical Trial Registry, ITMCTR2026000549.

04.
arXiv (CS.CV) 2026-06-11

Q-Fold: Query-Aware Focus-Context Spatio-Temporal Folding for Long Video Understanding

Long-video understanding remains challenging for multimodal large language models, because temporally extended videos often contain thousands of frames and are therefore expensive to process exhaustively. Existing methods usually construct compact visual inputs from long videos under a limited visual budget. However, most of them still follow a frame-centric paradigm and apply similar representations to retained content regardless of its importance. This makes it difficult to preserve both high-fidelity visual evidence and broad temporal coverage. To address this issue, we propose Q-Fold, a training-free input construction framework for long-video understanding. Instead of treating isolated frames as the basic modeling unit, Q-Fold operates on contiguous temporal segments and constructs a heterogeneous Focus–Context representation under query guidance. Query-relevant segments are preserved as high-fidelity Focus Frames, while less relevant segments are folded into chronology-preserving contextual layouts. In this way, Q-Fold preserves critical visual evidence and broad temporal coverage, while better maintaining local temporal continuity within short segments. Experiments on four long-video benchmarks with multiple Video-MLLMs show that Q-Fold consistently improves performance without increasing the input budget. Notably, it achieves gains of up to 9.1 percentage points on an ultra-long video benchmark. Code will be made publicly available.

05.
arXiv (CS.CL) 2026-06-17

Perceptual compensation for tonal context in self-supervised speech models

This study examines the extent to which the wav2vec2.0 architecture exhibits evidence of compensation for phonological context. We conducted a pseudo-replication of a perceptional compensation experiment on Mandarin Chinese tones, and compared the embedding similarities and probing classifier outputs between a purely self-supervised pre-trained model and a model fine-tuned for Mandarin ASR. No evidence of compensation was found in the embedding similarities of the purely pre-trained model. Probing classifiers showed some evidence of compensation in addition to the expected layer-wise improvements in categorization, but failed to replicate human performance on isolated test syllables. Our findings contrast with previous reports of sensitivity to phonological structure emerging through pre-training alone, and suggest that supervised objectives may be necessary to encourage the abstraction of at least some types of phonological regularities.

06.
arXiv (CS.AI) 2026-06-11

Are Frontier LLMs Ready for Cybersecurity? Evidence for Vertical Foundation Models from Dual-Mode Vulnerability Benchmarks

arXiv:2605.23243v2 Announce Type: replace-cross Abstract: We evaluate whether frontier LLMs are ready for cybersecurity through a dual-mode benchmark: white-box function-level vulnerability detection (VulnLLM-R, across C/Java/Python) and black-box web application security testing (five production-style applications with 118 ground-truth vulnerabilities across 20+ CWE families, which we will open-source). We test six frontier models (GPT-5.4, Codex~5.3, Claude Opus~4.6, Sonnet~4.6, Gemini~3.1~Pro and Gemini~3~Flash) and two domain-specialized models across four testing paradigms. Our findings are sobering: (1)~every frontier model produces 10-50% false positive rates in white-box detection, systematically over-predicting vulnerabilities; (2)~in black-box testing, frontier models achieve only 4-8% ground-truth coverage, improving to just 10-19% even with external security tools (Playwright MCP, Burp Suite MCP); (3)~structured penetration-testing methodology encoded in domain-specialized agents raises per-family detection above 50%, demonstrating that methodology, not scale, is the primary lever; and (4)~a domain-specialized defense model achieves the highest precision (0.904) and lowest false positive rate (9.7%) among all models, on a single GPU. We identify the absence of structured security testing traces end-to-end request/response sequences, failure-heavy data, and multi-step attack chains as the fundamental training data bottleneck, and propose self-play security testing as a data generation strategy. Our results make the case for vertical foundation models purpose-built for cybersecurity.

07.
arXiv (CS.CV) 2026-06-17

R1-SyntheticVL: Is Synthetic Data from Generative Models Ready for Multimodal Large Language Model?

In this work, we aim to develop effective data synthesis techniques that autonomously synthesize multimodal training data for enhancing MLLMs in solving complex real-world tasks. To this end, we propose Collective Adversarial Data Synthesis (CADS), a novel and general approach to synthesize high-quality, diverse and challenging multimodal data for MLLMs. The core idea of CADS is to leverage collective intelligence to ensure high-quality and diverse generation, while exploring adversarial learning to synthesize challenging samples for effectively driving model improvement. Specifically, CADS operates with two cyclic phases, i.e., Collective Adversarial Data Generation (CAD-Generate) and Collective Adversarial Data Judgment (CAD-Judge). CAD-Generate leverages collective knowledge to jointly generate new and diverse multimodal data, while CAD-Judge collaboratively assesses the quality of synthesized data. In addition, CADS introduces an Adversarial Context Optimization mechanism to optimize the generation context to encourage challenging and high-value data generation. With CADS, we construct MMSynthetic-20K and train our model R1-SyntheticVL, which demonstrates superior performance on various benchmarks.

08.
arXiv (CS.AI) 2026-06-25

TokenMinds: Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems

arXiv:2606.25147v1 Announce Type: cross Abstract: User modeling in industrial recommender systems typically produces dense embeddings, which suffer from representational constraints inherent to fixed-dimensional vectors. An emerging alternative for discrete user representation – using LLMs to generate text-based user tokens – captures topical co-occurrences rather than deep sequential behavior dynamics and produces outputs that are difficult to ground to item attributes. Meanwhile, Semantic ID (SID) based item tokenization has proven effective for improving generalization in generative recommendation, yet discrete SID-based representations for users remain largely unexplored. We propose TokenMinds, an industrial-scale system that extends the PLUM framework from item retrieval to user modeling, generating both discrete SID-based user tokens and dense user embeddings via an encoder-decoder architecture adapted from pre-trained LLMs. This dual-output design provides the complementary benefits of discrete, semantically grounded user representations while maintaining compatibility with existing downstream models that rely on dense embeddings. Additionally, the shared SID vocabulary naturally extends to cross-scenario modeling: by unifying long-form and short-form video behaviors into a single model, we substantially reduce training and serving costs. We validate TokenMinds through extensive offline experiments and live launches on multiple YouTube surfaces, served on full user traffic (billions of users) via an asynchronous infrastructure that decouples representation generation from downstream scoring. Focusing on ranking as the primary downstream use case, our results confirm the practical viability of SID-based user tokens at industrial scale and demonstrate that tokens and dense embeddings provide complementary value across different production ranking systems.

09.
arXiv (CS.LG) 2026-06-17

Continuous-time Optimal Stopping through Deep Reinforcement Learning

arXiv:2606.17545v1 Announce Type: new Abstract: Simulation based solvers for optimal stopping problems must discretize the stopping decision. Under classical dynamic programming, a coarse exercise grid with only a few stopping opportunities can materially undervalue the optimal expected reward, whereas on a very fine grid, approximation errors accumulate through the backward recursion. To remove this limitation, we develop a new reinforcement-learning inspired algorithm that enables us to learn the exercise rule at arbitrarily fine time resolution. Our CARLOS (Continuous-time Adaptive Reinforcement Learning for Optimal Stopping) algorithm utilizes an aggregate deep neural network (ADNN) to learn a joint space-time decision boundary. Starting from a coarse time grid, we progressively increase the frequency of stopping opportunities, while in parallel training the ADNN to refine its timing-value estimates. We moreover design an adaptive sampling strategy that gradually concentrates training effort near the stopping boundary. Benchmarked results show that CARLOS delivers higher prices than existing Bermudan solvers, approaching the American upper bound, and achieves high computational efficiency relative to non-RL comparators.

10.
arXiv (quant-ph) 2026-06-24

Optimizing LOCC Protocols on Product Stiefel Manifold

arXiv:2510.06909v2 Announce Type: replace Abstract: Characterizing the operational limits of Local Operations and Classical Communication (LOCC) is a central problem in distributed quantum information, yet remains computationally intractable due to the non-convex geometry of the LOCC set. We introduce a geometric framework that embeds the physical constraints of fixed-round LOCC protocols onto the product Stiefel manifold, converting a constrained protocol-design problem into unconstrained Riemannian optimization. We demonstrate this framework through entanglement distillation: by directly optimizing finite-copy LOCC protocols, we discover achievable protocols whose fidelities match positive partial transpose (PPT) upper bounds to within numerical precision, and we provide numerical evidence for both the operational advantage of adaptive communication rounds and the super-additivity of coherent information under two-way processing. These results establish Riemannian manifold optimization as a practical tool for probing the physical limits of future quantum networks.

11.
medRxiv (Medicine) 2026-06-15

Reaching out-of-school girls with HPV vaccination: A qualitative evaluation in six low- and middle-income countries using the RE-AIM framework

Background Infection with human papillomavirus (HPV), the primary cause of cervical cancer, disproportionately affects women in low- and middle-income countries (LMICs). While school-based vaccination of adolescent girls against HPV is highly effective, this strategy systematically excludes out-of-school (OOS) girls. Using the RE-AIM framework, we explored strategies to reach OOS girls with HPV vaccination across six African and Asian LMICs. Methods We conducted semi-structured key informant interviews with 32 vaccination program stakeholders from Cambodia, Cameroon, Kenya, Malawi, Mozambique, and Uganda between May and September 2024. Interviews explored countries implementation successes, challenges, and strategies to reach OOS girls with HPV vaccination and sustainability considerations. Data were analyzed using a hybrid team-based thematic analysis approach guided by the RE-AIM framework. Results Community outreach-based strategies, typically integrated into routine immunization outreach, were identified as the most effective approach to reach OOS girls with HPV vaccination. Targeted strategies, such as locating outreach clinics in community venues frequented by OOS girls (e.g., churches, markets) enhanced implementation. Perceived effectiveness of these strategies varied across participants, and formal assessment of effectiveness was constrained by the absence of disaggregated vaccination coverage data by school enrollment status. Some subpopulations of OOS girls (i.e., girls in nomadic or migrant communities, urban OOS girls) were not readily reached through standard outreach approaches, prompting implementation of adapted and tailored strategies for these subpopulations. Costs associated with conducting outreach in harder-to-reach areas were major barriers to reaching OOS girls, presenting challenges to the sustainability and cost-effectiveness of these approaches. Conclusions Routine community outreach platforms were widely perceived as most effective for reaching OOS girls. Strengthening disaggregated monitoring systems, adapting outreach for harder-to-reach subpopulations of OOS girls, and financing delivery models for tailored outreach strategies will be critical to improving equitable HPV vaccine coverage among OOS girls.

12.
arXiv (quant-ph) 2026-06-25

Hodge Spectral Surrogates for Topology-Constrained Optimization

arXiv:2606.25194v1 Announce Type: cross Abstract: Topological information is widely used in data analysis, network design, and machine learning, and topological constraints naturally arise when optimizing or generating objects with prescribed homological structure. However, directly controlling Betti numbers and persistent homology is difficult because they are discrete and combinatorial. We propose a differentiable framework for topology-constrained optimization based on Hodge-spectral relaxations of homological constraints and low-pass spectral filters. From soft graphs and soft clique complexes, we construct Hodge-Laplacian-type spectral relaxations that unify graph clique complexes and Vietoris–Rips filtrations of point clouds. In the hard limit, the penalty-regularized ambient operator recovers the ordinary Hodge Laplacian on the active subcomplex, while in the soft regime it serves as a differentiable low-frequency spectral surrogate. Homological information is represented by zero and near-zero modes, and differentiable topological objectives are defined using heat filters, resolvent filters, and polynomial Laplacian moments. For point clouds, we show that the proposed Hodge spectral-filter losses yield more spatially distributed gradients, smoother scale-normalized behavior under persistence-pairing changes, and geometry-aware update directions than persistent-homology-based losses. For graph clique complexes, Laplacian moments control normalized first-Betti-type quantities and can be combined with ordinary graph-feature objectives. We also discuss connections to trace-based normalized Betti-number estimation, polynomial spectral methods, and possible quantum trace estimation.

13.
arXiv (CS.LG) 2026-06-18

Automated Byzantine-Resilient Clustered Decentralized Federated Learning for Battery Intelligence in Connected EVs

arXiv:2605.21115v2 Announce Type: replace-cross Abstract: Federated learning (FL) has emerged as a promising paradigm for managing electric vehicle (EV) battery data in intelligent transportation systems (ITS), enabling privacy-preserving tasks such as anomaly detection and capacity estimation. However, most existing frameworks rely on centralized aggregation schemes, which pose critical limitations in terms of security and trust. To address these challenges, we propose ABC-DFL, an automated Byzantine-resilient clustered decentralized federated learning (C-DFL) framework for connected EVs. The proposed incentive-driven C-DFL system replaces the central server with an open-permissioned blockchain, featuring a new dynamic Quorum Byzantine Fault Tolerance (QBFT) protocol and an oracle-based aggregation layer, to enhance trust, security, and automation. At the core of ABC-DFL lies FLECA (Filtered Layered Enhanced Clustering Aggregation), a robust hierarchical aggregation protocol that mitigates Byzantine attacks by having each EV filter malicious updates using an adaptive threshold based on deviations from its reference model update. Oracle nodes, responsible for inter-group aggregation, employ robust clustering to isolate and aggregate model updates from trustworthy EV groups. Comprehensive experimental evaluations demonstrate that FLECA matches FedProx convergence under benign conditions and significantly outperforms existing defenses with attack impact scores below 0.10 in adaptive adversarial scenarios. Furthermore, several learning experiments with multitask models confirm the effectiveness and fairness of the incentive mechanism. Finally, on-chain and off-chain benchmarks validate the practicality of ABC-DFL.

14.
arXiv (CS.CL) 2026-06-12

Beyond the Commitment Boundary: Probing Epiphenomenal Chain-of-Thought in Large Reasoning Models

Chain-of-thought (CoT) reasoning is the dominant paradigm for inference-time scaling in language models, yet the causal influence of individual steps on the final answer poorly understood. We estimate each step's causal importance via early exit and use this measure to study how answers form across the reasoning traces of several model families. Across diverse tasks, we find that reasoning typically crosses a commitment boundary – a sharp transition from transient intermediate guesses to a stable, high-confidence answer. This transition often happens in a single step, well before the model's reasoning block ends, and is followed by epiphenomenal CoT steps that leave the final answer probability unaltered. Using attention probes, we show that answer-formation stages can be linearly decoded from intermediate reasoning steps with high accuracy and generalize robustly to unseen reasoning tasks. We exploit this signal to early-exit reasoning blocks at the commitment boundary, reducing the length of CoTs up to 55\% on average with negligible impact on model performance.

15.
arXiv (CS.LG) 2026-06-16

GD$^2$PO: Mitigating Multi-Reward Conflicts via Group-Dynamic reward-Decoupled Policy Optimization

arXiv:2606.16771v1 Announce Type: new Abstract: As LLMs advance, post-training reinforcement learning (RL) increasingly relies on multi-dimensional rewards to cultivate comprehensive capabilities. This shift demands new algorithms capable of optimizing diverse and potentially competing objectives simultaneously. To address this, existing methods such as Group reward-Decoupled Policy Optimization (GDPO) decompose the overall score into independent reward groups, then compute the RL loss separately within each group. However, this strategy still encounters multi-reward conflicts: a single rollout can yield positive advantages on certain reward dimensions but negative ones on others, causing opposing signals to cancel each other out during aggregation, further hindering RL training efficiency. Inspired by Dynamic sAmpling Policy Optimization (DAPO), which improves RL training efficiency by filtering out ineffective rollouts with near-zero advantages, we propose Group-Dynamic reward-Decoupled Policy Optimization (GD$^2$PO). Specifically, GD$^2$PO employs a conflict-aware filtering mechanism to mask out rollouts suffering from severe reward-wise disagreement. By preventing conflicting signals from canceling each other out, this masking strategy preserves and enhances the magnitude of effective RL advantages, thereby significantly accelerating learning efficiency. Furthermore, we introduce query-level reweighting to dynamically adjust the update intensity of each query based on its overall reward consensus. Experiments on various multi-reward scenarios, including tool calling and human preference alignment, demonstrate that GD$^2$PO consistently and significantly outperforms existing baselines. The code is available at https://github.com/Qwen-Applications/GD2PO.

16.
medRxiv (Medicine) 2026-06-24

Developing and Evaluating an Online Educational Program for Falls Prevention Care in Community Optometric Primary Care Settings: A Pilot Study

Introduction Globally, falls are the leading cause of injury hospitalisation, with vision being a significant falls risk factor. Community optometrists, as primary eye care professionals, are well positioned to contribute to falls prevention care. However, scant studies have evaluated if education could enable optometrists to incorporate falls prevention care into practice. This two-phase pilot study aimed to design and develop an online education program for community optometrists to deliver primary falls prevention care and to evaluate optometrists reaction to, and learning from, the education. Methods In phase one, an education program was designed by optometrists and falls experts and published online. In phase two, community optometrists were recruited through convenience sampling to undertake the education. Guided by the New World Kirkpatrick model(R) of training evaluation, reaction and learning were evaluated using pre/post surveys. Quantitative data were analysed using Wilcoxon sign-rank tests and McNemar Exact Tests and qualitative responses using inductive content analysis. Results Participants (n=13) reported high levels of satisfaction and engagement with the online education and unanimously endorsed its relevance to clinical practice. Participants demonstrated significantly improved knowledge and awareness of falls prevention post-education, compared to pre-education and were significantly more confident to enact falls prevention care. Perceived enablers to providing falls prevention care included having access to practical resources and ongoing education. Time constraints during consultation and cost to patients for further care if subsequent referrals were made were identified as possible barriers to providing falls prevention care. Conclusion Online education improved community optometrists knowledge and confidence to provide falls prevention care. Further research that evaluates the effectiveness of continuing education for optometrists to enact falls prevention care into practice is required.

17.
arXiv (CS.AI) 2026-06-16

OSGuard: A Benchmark for Safety in Computer-Use Agents

arXiv:2606.15034v1 Announce Type: new Abstract: Computer-use agents are increasingly evaluated by whether they complete realistic desktop and web tasks. However, task success alone can miss failures in which an agent reaches the nominal goal through an unsafe shortcut. We introduce OSGuard, a dual-granularity benchmark suite for evaluating safety in computer-use agents under benign, unchanged user instructions. OSGuard contains an action-level benchmark for local guardrail decisions and a risk-augmented execution suite for end-to-end evaluation. The action-level benchmark consists of contextualized proposed actions labeled as allowed, unrelated, or unsafe, each judged relative to the original instruction and current interface state. The execution suite contains manually constructed OSWorld-derived task variants in which the original task remains achievable, but the environment is modified to introduce latent hazards such as destructive overwrites, etc. Each variant is paired with augmented evaluators that retain the original task-success criterion while adding explicit state-based safety invariants, allowing us to distinguish safe completions from unsafe completions that satisfy the nominal task objective. Our experimental results on OSGuard show that current multimodal guardrails can perform well on isolated action judgments, while risk-augmented execution exposes remaining gaps between local oversight and reliable end-to-end safety. This dual-granularity design enables more precise diagnosis of whether models can both recognize unsafe proposed actions and improve full-task safety when deployed as guardrails.

18.
arXiv (CS.CL) 2026-06-25

ConPress: Learning Efficient Reasoning from Multi-Question Contextual Pressure

Large reasoning models (LRMs) typically solve reasoning-intensive tasks by generating long chain-of-thought (CoT) traces, leading to substantial inference overhead. We identify a reproducible inference-time phenomenon, termed Self-Compression: when multiple independent and answerable questions are presented within a single prompt, the model spontaneously produces shorter reasoning traces for each question. This phenomenon arises from multi-question contextual pressure during generation and consistently manifests across models and benchmarks. Building on this observation, we propose ConPress (Learning from Contextual Pressure), a lightweight self-supervised fine-tuning approach. ConPress constructs multi-question prompts to induce self-compression, samples the resulting model outputs, and parses and filters per-question traces to obtain concise yet correct reasoning trajectories. These trajectories are directly used for supervised fine-tuning, internalizing compressed reasoning behavior in single-question settings without external teachers, manual pruning, or reinforcement learning. With only 8k fine-tuning examples, ConPress reduces reasoning token usage by 59% on MATH500 and 33% on AIME25, while maintaining competitive accuracy.

19.
arXiv (CS.CL) 2026-06-18

PatchWorld: Gradient-Free Optimization of Executable World Models

Text-agent environments are typically modeled as partially observable Markov decision processes (POMDPs), assuming that the simulator's latent state and transition dynamics are hidden from the agent. Yet little work has examined whether executable code can be induced to serve as a world model for prediction and planning under partial observability. We introduce PatchWorld, a gradient-free framework that turns offline trajectories into executable Python world models through counterexample-guided code repair. Instead of predicting the next observation with a black-box model, PatchWorld induces symbolic belief-state programs whose action updates can be inspected, replayed, and locally patched. Across seven AgentGym environments, PatchWorld-Simple achieves the highest code-based planning score among evaluated methods, reaching 76.4\% macro success in live one-step lookahead while invoking no LLM calls inside the world-model prediction module itself. We further find that a human-specified residual-memory bias improves surface observation fidelity but weakens decision utility. This exposes a tradeoff in executable world models, since improving observation fidelity can come at the expense of action-discriminative dynamics, and vice versa. Code is available at https://github.com/HKBU-KnowComp/PatchWorld.

20.
arXiv (CS.AI) 2026-06-15

The Weight Norm Sets the Grokking Timescale: A Causal Delay Law

arXiv:2606.13753v1 Announce Type: cross Abstract: Grokking is the delayed onset of generalization in neural networks, arising long after they fit the training data. Whether the weight norm causes this delay is disputed: some studies report a critical norm at the transition, others observe grokking with no fixed norm at all. We settle this by intervening on the norm during training rather than only observing it. Under free training with weight decay, networks grok when the weight norm reaches a value Wc that varies little across seeds and learning rates (CV 1 to 2 percent) and grows with the modular base as a power law. When we instead clamp the norm to a fixed multiple rho of Wc and hold it there, the network still groks, but the delay follows T_grok proportional to exp(alpha rho). One exponent, alpha near 7.5, fits this delay across four moduli (R^2 = 0.996). Over the swept ranges the held norm moves the delay by about 19x and the learning rate by only about 2x, and holding the norm above Wc slows grokking rather than preventing it. A final LayerNorm removes the dependence by decoupling weight scale from the network function; without it the exponential law returns. This pinned-norm delay is the exponential counterpart to the logarithmic delay predicted for a freely contracting norm.

21.
arXiv (CS.AI) 2026-06-24

DynaWM: Dynamics-Aware Distillation with World Model and Momentum Targets for Smooth Locomotion over Continuous Stairs

arXiv:2606.24089v1 Announce Type: cross Abstract: Recent advances in control have enabled bipedal-wheeled robots to traverse slopes and single-step obstacles, yet long staircase traversal remains challenging as current teacher-student frameworks suffer from weakened dynamics-aware representations and incomplete terrain geometry encoding. To bridge this gap, we propose DynaWM, a dynamics-aware representation learning framework. To enhance terrain encoding capability and enable transparent assessment, we introduce a world model as a regularizer to enforce forward-dynamics awareness, preserving comprehensive terrain geometry while facilitating hierarchical encoding visualization. To stabilize knowledge transfer, we employ a momentum target encoder to provide consistent distillation targets, preventing dimensional collapse from non-stationary teacher updates. Evaluation of the learned representations through Principal Component Analysis (PCA) visualization and quantitative metrics reveals that our encoder hierarchically captures terrain geometry with higher terrain encoding capability, leading to enhanced terrain adaptability and motion smoothness. Experimental results in simulation and real hardware demonstrate that our method achieves superior terrain adaptability and motion smoothness, enabling bipedal-wheeled robots to overcome diverse continuous stairs, as shown in Fig. 1.

22.
arXiv (CS.AI) 2026-06-12

The Emergence of Autonomous Penetration Capabilities in Large Language Model-Powered AI Systems

arXiv:2606.13079v1 Announce Type: cross Abstract: Nowadays, the autonomous execution of cyberattacks capable of causing substantial real-world harm is widely regarded as one of the critical red lines that frontier AI systems must not cross. Within this broader red-line scenario, autonomous penetration represents a core enabling capability and subtask: the ability of LLM-powered AI systems to independently conduct adversarial operations against a target server without human intervention, identify and exploit vulnerabilities, and obtain unauthorized access or control. A growing body of work has sought to assess the autonomous penetration capabilities of AI systems. However, existing evaluations often employ opaque methodologies, rely on unrealistic or overly simplified penetration-testing scenarios, or provide LLMs with excessive prior knowledge and task-specific guidance, and cannot accurately capture the extent to which modern AI systems can autonomously perform this core capability within broader high-impact cyberattack scenarios. To address these limitations, we construct a new autonomous penetration evaluation framework consisting of two components: target servers and agent scaffolding. Specifically, on the target-server side, we design two levels of target environments based on the number of secure services without known vulnerabilities deployed alongside a vulnerable service: Tier~1 (one secure service) and Tier~2 (three secure services), resulting in a total of 300 target servers. Meanwhile, the agent scaffolding adopts a general-purpose agent architecture equipped with a set of general-purpose cybersecurity tools, without any target-specific prior knowledge. We evaluate 19 open-weight and proprietary LLMs, and find that current models achieve penetration success rates ranging from 10.7% to 69.3%. Moreover, we observe that autonomous penetration capability continues to improve alongside advances in overall model capability.

23.
arXiv (CS.CL) 2026-06-16

Exploring Extrinsic and Intrinsic Properties for Effective Reasoning with Code Interpreter

Reasoning with a Code Interpreter (CI) has emerged as an effective paradigm for enhancing the reasoning capabilities of large language models (LLMs) through executable computation and iterative verification. Despite its growing adoption, the behavioral properties underlying effective code reasoning remain largely underexplored. In this work, we investigate code reasoning from two distinct perspectives inspired by prior studies of natural language reasoning: extrinsic properties, represented by crucial tokens, and intrinsic properties, represented by code-specific cognitive behaviors. Across multiple LLMs, we find that stronger CI reasoning models consistently exhibit a higher prevalence of crucial tokens and cognitive behaviors, particularly verification, backtracking, and backward chaining. Building on these observations, we examine how these properties can be leveraged during both inference and training. At inference time, appending code-specific crucial tokens improves performance on several reasoning capabilities, including mathematical, ordering, and optimization, while yielding limited benefits elsewhere. At training time, augmenting a state-of-the-art framework with code-specific cognitive behaviors improves supervised fine-tuning and reinforcement learning performance in two of three evaluated models. Further analysis shows that these behaviors reduce overthinking in incorrect responses and improve token efficiency, while also revealing factors that limit gains in a certain model. Our findings provide the first systematic characterization of effective reasoning with CI and demonstrate both the potential and limitations of leveraging key properties to improve CI-based reasoning.

24.
Nature (Science) 2026-06-10

Improved quantum processor logical error rates via correction and detection

作者:

Performing quantum algorithms for critical problems in physics and chemistry requires substantially lower error rates than the physical error rates of present quantum computers. Achieving such low logical error rates requires quantum error correction1,2 and physical error rates below a critical threshold value3–8. We experimentally demonstrate on a trapped-ion quantum charge-coupled device (QCCD)9,10 improvements in logical error rates ranging from 11× to 800× compared with several physical circuit baselines, including quantum computation on multiple qubits. Our results hinge on two quantum error correction code constructions optimized for an ion-trap processor: a 12-qubit code encoding two qubits inspired by Knill11 and a 16-qubit tesseract colour code encoding four qubits12,13. These constructions are combined with a scalable method of error detection and post-selection to achieve reduced logical error rates. Our results show that state-of-the-art quantum devices are already able to make use of fault tolerance and error correction to strongly suppress errors in non-trivial quantum circuit computations. Experimental demonstration of quantum error-correcting codes combined with error detection and post-selection applied to a trapped-ion quantum processor shows improvements in logical error rates ranging from 11× to 800× compared with several physical circuit baselines.

25.
arXiv (CS.CL) 2026-06-18

The Wrong Kind of Right: Quantifying and Localizing Misfired Alignment in LLMs

Warning: This paper studies stereotypes and biases, and contains potentially disturbing examples, used for illustration purposes only. Our findings should not be interpreted as an argument against alignment. Instead, this paper highlights the need for principled approaches to more advanced alignment. Alignment aims to ensure that large language models (LLMs) behave safely and reliably, including by avoiding unsafe inferences. However, we show that such safety-oriented behaviors can misfire: models may reject warranted conclusions even when they are explicitly supported by context. We call this failure mode misfired alignment, where alignment-induced changes cause LLMs to override explicit evidence. To quantify this phenomenon, specifically on stereotype-related alignment, we introduce VETO, a benchmark consisting of 2,032 BBQ-derived contrastive pairs, and define a new metric, Misfired Alignment Rate (MAR), which measures on a 0 to 100 scale how often a model fails on a stereotype-related question but succeeds on its contrastive counterpart. We benchmark 25 LLMs on VETO, and show that all LLMs, including the most recent ones, exhibit non-trivial (4.7 to 18.9%) MARs while all human participants achieve 0.0% MAR. Controlled priming experiments further show that alignment-induced cues can substantially amplify MAR across LLMs, indicating that these failures are not merely artifacts of individual examples but can be induced by safety-related framing. Mechanistic analyses on open-weight LLMs reveal late-layer suppression of evidence-supported answers, and comparisons between instruct and base LLMs suggest that this suppression emerges after instruction training. These findings show that current alignment methods can overgeneralize surface-level safety cues, to the point of overriding objective evidence, motivating more work on alignment objectives that better preserve contextual grounding.