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.
arXiv (quant-ph) 2026-06-11

Entanglement preservation and Clauser-Horne nonlocality in electromagnetically induced transparency quantum memories

arXiv:2507.15453v4 Announce Type: replace Abstract: Entanglement preservation in noisy quantum memories represents a central challenge in quantum information science. While experiments have shown that electromagnetically induced transparency (EIT) memories can store entangled photons, a quantitative theoretical analysis of whether nonlocal quantum correlations can survive storage loss induced by ground-state decoherence remains limited. Here we combine the dark-state polariton formalism with a reduced density-operator treatment to derive an EIT-specific effective pure-loss description for the retrieved photonic state in the ground-state-decoherence-limited regime. The analysis reveals that decoherence transforms an initially pure Bell state into a mixed state with a vacuum component and predicts a protocol-dependent storage-efficiency benchmark of 89.7% for violating the chosen unconditional Clauser-Horne (CH) inequality. Above this benchmark, the retrieved photonic state violates the CH inequality without post-selection, whereas below it, this unconditional CH violation is no longer obtained. This framework provides a quantitative theoretical description of entanglement retention, retrieved photonic density operators, and protocol-dependent Bell-test benchmarks in EIT quantum memories.

02.
arXiv (CS.CL) 2026-06-16

S1-DeepResearch: Beyond Search, Toward Real-World Long-Horizon Research Agents

Deep research agents aim to solve complex knowledge-intensive tasks through long-horizon planning, evidence gathering, reasoning, and report generation. While recent progress in search agents has demonstrated strong capabilities in information retrieval and answer verification, most existing training datasets remain search-centric, focusing primarily on closed-ended question answering and information localization. As a result, they mainly train information-seeking behavior while providing limited coverage of key deep research capabilities, including evidence integration, knowledge synthesis, planning, file understanding, and structured report generation. In this work, we propose a unified trajectory construction paradigm for deep research agents that combines closed-ended QA and open-ended exploration. The proposed framework consists of graph-grounded task formulation, agentic trajectory rollout, and multi-dimensional trajectory verification, enabling scalable synthesis of high-quality agentic trajectories spanning long-chain complex reasoning, deep research instruction following, report writing, file understanding and generation, and skills usage. Compared with existing search-oriented datasets, our synthesized trajectories place greater emphasis on knowledge synthesis, complex reasoning, and planning. S1-DeepResearch-32B achieves state-of-the-art performance among open-source models of comparable scale across 20 benchmarks spanning five capability dimensions, including complex reasoning, instruction following, report generation, file understanding, and skills usage. On several challenging deep research benchmarks, it approaches the performance of leading proprietary frontier models. These results highlight the importance of jointly modeling information acquisition, knowledge synthesis, and planning-oriented agent behaviors for building effective deep research agents.

03.
arXiv (CS.CL) 2026-06-15

Did You Forget What I Asked? Prospective Memory Failures in Large Language Models

作者:

Large language models often fail to satisfy formatting instructions when they must simultaneously perform demanding tasks. We study this behaviour through a prospective memory inspired lens from cognitive psychology, using a controlled paradigm that combines verifiable formatting constraints with benchmark tasks of increasing complexity. Across three model families and over 8,000 prompts, compliance drops by 2-21% under concurrent task load. Vulnerability is highly type-dependent: terminal constraints (requiring action at the response boundary) degrade most, with drops up to 50%, while avoidance constraints remain comparatively robust. A salience-enhanced format (explicit instruction framing plus a trailing reminder) recovers much of the lost compliance, restoring performance to 90-100% in many settings. Interference is bidirectional: formatting constraints can also reduce task accuracy, with one model's GSM8K accuracy dropping from 93% to 27%. In additional stacking experiments, joint compliance declines sharply as constraints accumulate. All results use deterministic programmatic checkers without an LLM-as-judge component on publicly available datasets.

04.
arXiv (quant-ph) 2026-06-15

Quantitative and Optimal Device-Independent Lower Bounds on Detection Efficiency

arXiv:2511.19302v2 Announce Type: replace Abstract: This paper examines a quantitative and optimal lower bound on the detector efficiency in a (2,2,2) Bell experiment within a fully device-independent framework, whereby the detectors used in the experiment are uncharacterized. We provide a tight lower bound on the minimum efficiency required to observe a desired Bell-CHSH violation using the Navascués-Pironio-Acín (NPA) hierarchy, confirming tightness up to four decimal places with numerical optimization over explicit quantum realizations. We then introduce the effect of dark counts and demonstrate how to quantify the minimum required efficiency to observe a desired CHSH violation with an increasing dark count error. Finally, to obtain an analytical closed-form expression of the minimum efficiency, we consider the set of no-signaling behaviors that satisfy the Tsirelson bound, which are easier to characterize than the quantum set. Using such behaviors, we find a simple closed-form expression for a lower bound on the minimum efficiency which is monotonically increasing with the CHSH violation, though the analytically obtained lower bounds are meaningfully below the numerically tight lower bound.

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

Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond

arXiv:2604.22748v3 Announce Type: replace Abstract: As AI systems move from generating text to accomplishing goals through sustained interaction, the ability to model environment dynamics becomes a central bottleneck. Agents that manipulate objects, navigate software, coordinate with others, or design experiments require predictive environment models, yet the term world model carries different meanings across research communities. We introduce a "levels x laws" taxonomy organized along two axes. The first defines three capability levels: L1 Predictor, which learns one-step local transition operators; L2 Simulator, which composes them into multi-step, action-conditioned rollouts that respect domain laws; and L3 Evolver, which autonomously revises its own model when predictions fail against new evidence. The second identifies four governing-law regimes: physical, digital, social, and scientific. These regimes determine what constraints a world model must satisfy and where it is most likely to fail. Using this framework, we synthesize over 400 works and summarize more than 100 representative systems spanning model-based reinforcement learning, video generation, web and GUI agents, multi-agent social simulation, and AI-driven scientific discovery. We analyze methods, failure modes, and evaluation practices across level-regime pairs, propose decision-centric evaluation principles and a minimal reproducible evaluation package, and outline architectural guidance, open problems, and governance challenges. The resulting roadmap connects previously isolated communities and charts a path from passive next-step prediction toward world models that can simulate, and ultimately reshape, the environments in which agents operate. Code and resources are available at: https://github.com/matrix-agent/awesome-agentic-world-modeling.

06.
arXiv (CS.CV) 2026-06-18

E-VAds: An E-commerce Short Videos Understanding Benchmark for MLLMs

E-commerce short videos represent a high-revenue segment of the online video industry characterized by a goal-driven format and dense multi-modal signals. Current models often struggle with these videos because existing benchmarks focus primarily on general-purpose tasks and neglect the reasoning of commercial intent. In this work, we first propose a multi-modal information density assessment framework to quantify the complexity of this domain. Our evaluation reveals that e-commerce content exhibits substantially higher density across visual, audio, and textual modalities compared to mainstream datasets, establishing a more challenging frontier for video understanding. To address this gap, we introduce E-commerce Video Ads Benchmark, which is the first benchmark specifically designed for e-commerce short video understanding. We curated 3,961 high-quality videos from Taobao covering a wide range of product categories and used a multi-agent system to generate 19,785 open-ended Q&A pairs, which consist of five distinct tasks. Finally, we develop E-VAds-R1, an RL-based reasoning model featuring a multi-grained reward design called MG-GRPO. This strategy provides smooth guidance for early exploration while creating a non-linear incentive for expert-level precision. Experimental results demonstrate that E-VAds-R1 achieves a 109.2% performance gain in commercial intent reasoning with only a few hundred training samples. Data is available at https://github.com/TaobaoTmall-AlgorithmProducts/E-VAds_Benchmark.

07.
arXiv (CS.AI) 2026-06-18

MagpieTTS-LF: Inference-Time Long-Form Speech Generation Without Training on Long-Form data

arXiv:2606.18485v1 Announce Type: cross Abstract: Neural Text-to-Speech (TTS) systems achieve remarkable quality on short utterances but long-form speech generation shows prosodic drift, speaker inconsistencies and sentence boundary artifacts. Existing approaches either compress sequences, increase context length or naively concatenate independently synthesized chunks. We present an inference-time approach called MagpieTTS-LF that enables MagpieTTS to produce coherent long-form speech without model retraining. Our method introduces three key innovations: (1) soft attention priors to guide monotonic alignment while preserving past and future context; (2) a stateful inference algorithm that maintains context across sentence chunks, ensuring prosodic continuity; (3) history-aware text encoding that uses past text for discourse-level prosodic planning. Experiments on long texts show significant improvements in long-range intelligibility, prosodic coherence, speaker consistency, and boundary naturalness compared to other baselines.

08.
arXiv (CS.CV) 2026-06-17

Mordal: Automated Pretrained Model Selection for Vision Language Models

Incorporating multiple modalities into large language models (LLMs) is a powerful way to enhance their understanding of non-textual data, enabling them to perform multimodal tasks. Vision language models (VLMs) form the fastest growing category of multimodal models because of their many practical use cases, including in healthcare, robotics, and accessibility. Unfortunately, even though different VLMs in the literature demonstrate impressive visual capabilities in different benchmarks, they are handcrafted by human experts; there is no automated framework to create task-specific multimodal models. We introduce Mordal, an automated multimodal model search framework that efficiently finds the best VLM for a user-defined task without manual intervention. Mordal achieves this both by reducing the number of candidates to consider during the search process and by minimizing the time required to evaluate each remaining candidate. Our evaluation shows that Mordal can find the best VLM for a given problem using $8.9\times$–$11.6\times$ lower GPU hours than grid search. We have also discovered that Mordal achieves about 69\% higher weighted Kendall's $\tau$ on average than the state-of-the-art model selection method across diverse tasks.

09.
arXiv (CS.LG) 2026-06-19

Evaluating Universal Machine Learning Force Fields Against Experimental Measurements

arXiv:2508.05762v2 Announce Type: replace-cross Abstract: Universal machine learning force fields (UMLFFs) promise to revolutionize materials science by enabling rapid atomistic simulations across the periodic table. However, their evaluation has been limited to computational benchmarks that may not reflect real-world performance. We introduce UniFFBench, a comprehensive evaluation framework featuring the MinX dataset – a diverse collection of 1,500+ mineral systems spanning 85 elements, extreme thermodynamic conditions (0–5000 K, 0–1000 GPa), and structural complexity, including partial occupancy and disorder. This diversity, combined with experimental reference values for validation, enables assessment of UMLFF generalization across chemical space and conditions substantially beyond typical training scenarios. Our systematic evaluation of six state-of-the-art UMLFFs reveals a substantial ``reality gap'': models achieving impressive performance on computational benchmarks often fail when confronted with experimental complexity. Even the best-performing models exhibit higher density prediction error than the threshold required for practical applications. We observe disconnects between simulation stability and mechanical property accuracy, with prediction errors correlating with training data representation rather than the modeling method.

10.
arXiv (CS.AI) 2026-06-11

Subliminal Learning Is Steering Vector Distillation

arXiv:2606.00995v3 Announce Type: replace Abstract: Subliminal learning refers to a student language model acquiring a teacher's traits (e.g. a system-prompted preference for owls) when fine-tuned on the teacher's outputs, despite the outputs being semantically unrelated to those traits. It remains poorly understood how data without semantic meaning can transfer specific semantic traits. In this work, we show that subliminal learning is mediated by a single steering vector, i.e. a vector added to the model's activations. Across two open-source models, we find that the teacher's system prompt is well approximated by a steering vector, and that the student's behavior is driven by learning an aligned vector over fine-tuning. System prompts that are not well approximated by steering vectors are not subliminally learned. This is a special case of steering vector distillation, in which a student trained on the outputs of a steered teacher learns to imitate that steering. We demonstrate steering vector distillation on a range of semantic and random vectors. Adding a semantic vector to a model's activations can have both model-independent and model-specific (i.e. non-semantic) effects on its behavior, so generated data that is non-semantic can transmit a vector with semantic effects, enabling subliminal learning. This also explains why subliminal learning does not transfer between models. We find that adaptive optimizers are necessary for subliminal learning in language models: activation gradients on steered data carry a small but consistent component along the steering direction, and non-adaptive optimizers impede this by allowing outlier gradients to dominate.

11.
arXiv (CS.CV) 2026-06-12

Mana: Dexterous Manipulation of Articulated Tools

Articulated tool manipulation remains a major challenge in dexterous robotics due to the need to coordinate internal degrees of freedom and contact-rich interactions. While prior work has largely focused on rigid objects, articulated tool use remains underexplored because of its physical complexity and the difficulty of learning functional grasping and manipulation policies. We present Mana (Manipulation Animator), a general sim-to-real framework that reinterprets dexterous manipulation as an animation problem. Inspired by computer animation, Mana employs a coarse-to-fine pipeline that transforms procedurally-generated grasp keyframes into manipulation trajectories through motion planning and reinforcement learning. The data generation process is largely automatic, requiring only a few mouse clicks to specify functional affordances (

12.
arXiv (CS.CL) 2026-06-15

OdysSim: Building Foundation Models for Human Behavior Simulation

Large language models are increasingly deployed as human simulators for interactive evaluation and social simulation. Yet helpfulness-driven post-training pulls them toward a homogeneous, overly agreeable assistant register, creating a behavioral Sim2Real gap. We present OdysSim, the largest open systematic investigation of behavioral foundation models, i.e., models trained to simulate human behavior at scale. We propose SOUL, a taxonomy of five capability axes (CONV, SS, COG, ROLE, EVAL) that unifies 62 datasets and 23 benchmark tasks under one framework. Specifically, we curate the OdysSim corpus (21.4M interactions, 10B tokens, retrofitted with back-generated social contexts), construct the SOUL-Index benchmark, and develop an end-to-end training recipe combining midtraining, task-specific RL, and expert distillation. The resulting open 8B OSim model ranks first or tied-first on 8 of 23 tasks, outperforming any individual frontier model by this count, with the strongest gains on conversational and social tasks. Its outputs are also more human-like in length, formatting, and word choice, and it transfers zero-shot to out-of-distribution user simulation on $\tau$-bench, nearly matching real users on reaction alignment (93.2 vs. 93.5). We further show that LLM-as-judge RL induces reward-hacking patterns, and that our detectors can mitigate them during post-training. Together, our findings suggest that behavioral foundation models require rethinking the LLM training paradigm. We release all artifacts to support future research.

13.
arXiv (CS.CL) 2026-06-11

Calibration Drift Under Reasoning: How Chain-of-Thought Budgets Induce Overconfidence in Large Language Models

The ability of large language models (LLMs) to express calibrated uncertainty is important for safe deployment. Chain-of-thought (CoT) reasoning is widely used to improve accuracy and reliability, but its effect on calibration is not fully understood. We show that this picture is incomplete: in some settings, increasing the reasoning budget beyond a task-specific threshold can cause models to become systematically overconfident, assigning high confidence to incorrect answers. We call this phenomenon Calibration Drift Under Reasoning (CDUR) and study it both theoretically and empirically. We define reasoning budget B and analyze conditions under which Expected Calibration Error ECE(B) follows a non-monotonic pattern: it first decreases as reasoning corrects errors, then increases as longer reasoning produces internally consistent but incorrect explanations. We propose a Hypothesis Lock-In model based on autoregressive generation to explain this behavior. We evaluate Llama-3.1-8B and Llama-3.3-70B on 47 reasoning-trap questions across four reasoning budgets and three seeds (1,368 API calls; 574 valid responses). The 8B model shows non-monotonic calibration behavior, while results for the 70B model are limited to baseline evaluation and are inconclusive for budget-dependent effects. We introduce CABStop, a calibration-aware stopping rule that halts reasoning when confidence diverges from an auxiliary accuracy estimate. These results suggest that increasing reasoning depth does not always improve reliability and should be monitored carefully.

14.
arXiv (CS.CL) 2026-06-19

Pruning via Causal Attribution Preserves Reasoning Performance in Large Language Models

Large language models (LLMs) excel at multi-step reasoning but incur substantial inference cost. We introduce Causal Attribution Pruning (CAP), a training-free method that identifies critical attention heads by measuring their causal impact on reasoning tasks and uses these head-level scores to guide fine-grained weight pruning. For each attention head, CAP estimates the expected performance degradation when the head is masked during forward passes on a small calibration set of reasoning problems. These causal scores are then converted into weight-level importance values for the corresponding projection matrices. Unlike magnitude-only or activation-based criteria, CAP's interventional measurement directly captures each head's functional contribution, yielding relative accuracy gains of up to 61% over Wanda on ARC-Challenge at 20% sparsity. We evaluate CAP on GSM8K, StrategyQA, and ARC-Challenge using Llama-3-8B-Instruct and Mistral-7B-Instruct at 10%, 20%, and 50% sparsity. At moderate sparsity (10-20%), CAP improves over Wanda in most model-benchmark configurations. with especially large gains on ARC-Challenge for Llama-3. Our results suggest that attention-head-level causal attribution can better preserve reasoning performance on downstream benchmarks than correlational pruning criteria at equivalent sparsity, while remaining limited by coarse MLP attribution at 50% sparsity.

15.
medRxiv (Medicine) 2026-06-18

Age as a moderator of a brief alcohol intervention among injury patients in Northern Tanzania

Background: Alcohol use is a leading modifiable risk factor for injury in sub-Saharan Africa. In Tanzania, young people ([≤]24 years) experience greater alcohol-related harm despite drinking less frequently than adults. Punguza Pombe kwa Afya Yako (PPKAY) is a culturally adapted, brief intervention for injury patients in Tanzania. This study examined whether age moderates its effectiveness. Methods: We conducted an exploratory secondary analysis of baseline and 3-month data from the PPKAY randomized trial among injury patients aged [≥]18 years at Kilimanjaro Christian Medical Centre, Tanzania. Eligible participants reporting alcohol use before injury, AUDIT [≥]8, or positive breathalyzer were randomized to usual care or PPKAY with SMS boosters. The primary outcome was binge drinking days. Count outcomes were analyzed using negative binomial regression with robust SEs and continuous outcomes using mixed-effects models. Effect modification was assessed using a three-way interaction (Time x intervention x Age). Results: Among 543 participants (mean age 36.8 years; 16.2% aged 18–24), age moderated the intervention effect for drinking days (IRR = 0.27, 95% CI 0.07 – 0.98; p = 0.046) and drinks consumed (IRR = 0.17, 95% CI 0.04 – 0.77; p = 0.021). The intervention reduced 4 drinking days (95% CI -7.1 to -0.8) and 27.5 drinks (95% CI -42.8 to -12.2) among young people, while adults showed reductions in both arms, without intervention-specific effect. Conclusion: The effects of ED-based brief alcohol interventions are not uniform, varying across both age groups and alcohol-related outcomes. We found a greater responsiveness in drinking frequency and quantity reported among young people.

16.
arXiv (CS.LG) 2026-06-12

Estimating Individualized Treatment Effects in Acute Ischemic Stroke with Causal Transformation Models (TRAM-DAG): A Multi-Centre Observational Study with External RCT Validation

arXiv:2606.12623v1 Announce Type: cross Abstract: Personalized medicine in acute ischemic stroke requires moving beyond average treatment effects (ATE) to individualized treatment effect (ITE) estimates to support treatment decisions. In acute ischemic stroke, mechanical thrombectomy has been shown to be more effective on average than lysis in randomized controlled trials (RCTs), such as the MR CLEAN study. We aim to identify which individual patients benefit most from mechanical thrombectomy compared to lysis. The outcome of interest is the modified Rankin Scale (mRS) at three months, an ordinal measure of functional disability (0: no symptoms, 6: death). We demonstrate that causal transformation models on directed acyclic graphs (TRAM-DAG) can be used for ITE estimation after being fitted on observational MAGIC multi-center stroke patient data. To ensure comparability with the MR CLEAN population, which we use for validation, we train the TRAM-DAG on a MAGIC sub-population with NIHSS at admission >= 6, corresponding to one inclusion criterion of MR CLEAN. The fitted model is then used to estimate ITEs for stroke patients in the MR CLEAN population. While these ITE estimates cannot be confirmed experimentally, we show that their average is consistent with the trial's reported ATE. Furthermore, the ITE estimates correctly rank trial patients by their observed frequency of a good outcome (mRS at three months

17.
bioRxiv (Bioinfo) 2026-06-19

Children's DNA Methylation and Family Dynamics in a Congo Basin Subsistence Community: Links with Parental Conflict and Fathers' Caregiving

Family environments may contribute to children's long-term health through biological processes, including epigenetic regulation such as DNA methylation (DNAm). However, most studies in this area focus on Euro-American populations while also rarely including fathering data. The current study investigated children's blood DNAm associations with positive (father caregiving) and negative (parental conflict) family dynamics in a smaller-scale subsistence society living in the Congo Basin rainforest. We measured DNAm from dried blood spots of 54 children (mean age=8.48 years) and conducted three epigenome-wide association studies aimed at discovering differential co-methylated regions (CMRs) associated with family dynamics. Via path models, we investigated the health implications and shared contribution of family factors of the identified CMRs. Differential DNAm associated with family dynamics was localized to genes related to stress, immunology, development, and aging, thus possibly linking to children's physical health and were simultaneously connected to other family factors such as number of siblings. Our findings suggested similarities in biological embedding of family factors across socio-ecologically diverse contexts.

18.
arXiv (CS.LG) 2026-06-16

AREAL-DTA: Dynamic Tree Attention for Efficient Reinforcement Learning of Large Language Models

arXiv:2602.00482v2 Announce Type: replace Abstract: Reinforcement learning (RL)-based post-training for large language models (LLMs) is computationally expensive, as it generates many rollout sequences that frequently share long token prefixes. Existing RL frameworks usually process these sequences independently during policy training, i.e., repeatedly recomputing identical prefixes in both the forward and backward passes of policy gradient computation, leading to substantial inefficiencies in computation resources and memory usage. Although prefix sharing naturally induces a tree structure over rollouts, packed tree-mask approaches scale poorly in RL settings. In this paper, we introduce AReaL-DTA, which efficiently exploits prefix sharing in RL training. AReaL-DTA employs a depth-first search (DFS)-based execution strategy that dynamically traverses the rollout prefix tree during both forward and backward computation, materializing only a single root-to-leaf path at a time. To further improve scalability, AReaL-DTA incorporates a load-balanced distributed batching mechanism that dynamically constructs and processes prefix trees across multiple GPUs. On $\tau^2$-bench, AReaL-DTA improves training throughput by up to $8.31\times$ over dense training and up to $1.70\times$ over sparse training. Our code is available at https://github.com/areal-project/AReaL/tree/feat/dta.

19.
arXiv (CS.CL) 2026-06-11

LifeSentence: Language models can encode human life course trajectories from longitudinal panel data

Forecasting human life outcomes is important to gain insights into how individuals attain long and healthy lives. Conventional statistical approaches yield limited accuracy, potentially due to discarding the sequential structure of the life course. Modern methods such as transformer architectures require large scale training data that most longitudinal panel studies lack. Here we introduce LifeSentence, a model for life-course reasoning that bridges large language models with longitudinal panel data. By representing each life event as a structured natural-language record and instruction-tuning a pretrained 24-billion-parameter language model across an 18-task evaluation taxonomy spanning prediction, robustness and reasoning, LifeSentence supplements panel data with distributional knowledge already encoded during pretraining. Trained on approximately 65,000 individuals from the German Socio-Economic Panel - roughly 45 times fewer than prior transformer-based approaches - LifeSentence outperforms classical and deep learning baselines across all task families, achieving a threefold improvement in joint event-and-timing prediction from best baselines and 91.2% Kendall's tau when reconstructing chronological order from timestamp-stripped event sets. Without explicit supervision, the model recovers documented patterns of social stratification, including the education premium, the gender wage gap and the motherhood penalty, from discrete event sequences alone. A natural-language interface further enables qualitatively new research queries, such as connecting an early-life history to a specified late-life endpoint, establishing LifeSentence as both a predictive tool and a probe for counterfactual exploration of human biographies.

20.
arXiv (CS.AI) 2026-06-19

Bridging Distribution Shift and AI Safety: Conceptual and Methodological Synergies

arXiv:2505.22829v2 Announce Type: replace-cross Abstract: This paper bridges distribution shift and AI safety through a comprehensive analysis of their conceptual and methodological synergies. While prior discussions often focus on narrow cases or informal analogies, we establish two types connections between specific causes of distribution shift and fine-grained AI safety issues: (1) methods addressing a specific shift type can help achieve corresponding safety goals, or (2) certain shifts and safety issues can be formally reduced to each other, enabling mutual adaptation of their methods. Our findings provide a unified perspective that encourages deeper integration between distribution shift and AI safety research.

21.
arXiv (CS.CV) 2026-06-17

ThinkJEPA: Empowering Latent World Models with Large Vision-Language Reasoning Model

Recent progress in latent world models (e.g., V-JEPA2) has shown promising capability in forecasting future world states from video observations. Nevertheless, dense prediction from a short observation window limits temporal context and can bias predictors toward local, low-level extrapolation, making it difficult to capture long-horizon semantics and reducing downstream utility. Vision–language models (VLMs), in contrast, provide strong semantic grounding and general knowledge by reasoning over uniformly sampled frames, but they are not ideal as standalone dense predictors due to compute-driven sparse sampling, a language-output bottleneck that compresses fine-grained interaction states into text-oriented representations, and a data-regime mismatch when adapting to small action-conditioned datasets. We propose a VLM-guided JEPA-style latent world modeling framework that combines dense-frame dynamics modeling with long-horizon semantic guidance via a dual-temporal pathway: a dense JEPA branch for fine-grained motion and interaction cues, and a uniformly sampled VLM thinker branch with a larger temporal stride for knowledge-rich guidance. To transfer the VLM's progressive reasoning signals effectively, we introduce a hierarchical pyramid representation extraction module that aggregates multi-layer VLM representations into guidance features compatible with latent prediction. Experiments on hand-manipulation trajectory prediction show that our method outperforms both a strong VLM-only baseline and a JEPA-predictor baseline, and yields more robust long-horizon rollout behavior.

22.
arXiv (CS.AI) 2026-06-17

Breaking the Code: Security Assessment of AI Code Agents Through Systematic Jailbreaking Attacks

arXiv:2510.01359v2 Announce Type: replace-cross Abstract: Code-capable large language model (LLM) agents are embedded in software engineering workflows where they can read, write, and execute code, raising "jailbreak" stakes beyond text-only settings. Prior evaluations emphasize refusal or harmful-text detection, leaving open whether agents compile and run malicious programs. We present JAWS-Bench (Jailbreaks Across WorkSpaces), a benchmark spanning three escalating workspace regimes mirroring attacker capability: empty (JAWS-0), single-file (JAWS-1), and multi-file (JAWS-M). We pair this with a hierarchical, executable-aware Judge Framework that tests (i) compliance, (ii) attack success, (iii) syntactic correctness, and (iv) runtime executability, to measure deployable harm. Across seven LLM backends from five families, prompt-only attacks in JAWS-0 achieve 61% compliance; 58% are harmful, 52% parse, and 27% run end-to-end. In JAWS-1, compliance reaches ~100% for stronger models with a mean ASR (Attack Success Rate) ~71%; JAWS-M raises mean ASR to ~75%, with 32% runnable attack code. Wrapping an LLM in an agent increases ASR by 1.6$\times$, by overturning initial refusals during planning and tool use. Similar trends hold for OpenHands, SWE-Agent, and OpenAI Codex, suggesting our JAWS-Bench is agent-agnostic. Category analyses identify which attack classes are most vulnerable and deployable, motivating execution-aware defenses and refusal-preserving agent designs.

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

UniT: Unified Multimodal Chain-of-Thought Test-time Scaling

Unified models can handle both multimodal understanding and generation within a single architecture, yet they typically operate in a single pass without iteratively refining their outputs. Many multimodal tasks, especially those involving complex spatial compositions, multiple interacting objects, or evolving instructions, require decomposing instructions, verifying intermediate results, and making iterative corrections. While test-time scaling (TTS) has demonstrated that allocating additional inference compute for iterative reasoning substantially improves language model performance, extending this paradigm to unified multimodal models remains an open challenge. We introduce UniT, a framework for multimodal chain-of-thought test-time scaling that enables a single unified model to reason, verify, and refine across multiple rounds. UniT combines agentic data synthesis, unified model training, and flexible test-time inference to elicit cognitive behaviors including verification, subgoal decomposition, and content memory. Our key findings are: (1) unified models trained on short reasoning trajectories generalize to longer inference chains at test time; (2) sequential chain-of-thought reasoning provides a more scalable and compute-efficient TTS strategy than parallel sampling; (3) training on generation and editing trajectories improves out-of-distribution visual reasoning. These results establish multimodal test-time scaling as an effective paradigm for advancing both generation and understanding in unified models.

24.
arXiv (quant-ph) 2026-06-19

Truncated Wigner dynamics of biclique quantum spin glasses

作者:

arXiv:2606.20187v1 Announce Type: cross Abstract: Quantum spin glasses are often considered testbeds for studying quantum optimization algorithms and as such have been the subject of various quantum advantage claims. Here we investigate the near adiabatic dynamics of biclique quantum spin glasses within the (discrete) truncated Wigner approximation (TWA). Benchmarks on small systems show that TWA recovers sample-to-sample fluctuations of the Edwards-Anderson order parameter, over a wide range of annealing times, with increasing fidelity when the system size increases. We extract critical exponents from the Binder cumulant in line with theoretical expectations, reproducing recent quantum experiments. The computational cost of the method is minimal and it can easily be applied to tens of thousands of qubits.

25.
arXiv (CS.CV) 2026-06-16

Chronological Blindness: Benchmarking Temporal Reasoning in Vision-Language Models with CHRONOSIGHT

Human perception of visual scenes is inherently temporal. We instinctively recognise whether a fruit is ripening or rotting, whether construction is progressing or being demolished, and approximately how much time separates two photographs of the same subject. Whether large vision-language models (VLMs) share this competence remains an open and practically important question. We introduce CHRONOSIGHT, a rigorously controlled benchmark evaluating five dimensions of visual temporal reasoning: CHRONORANK (chronological ordering of image sequences), CHRONOLOCATE (ordinal stage localisation from a single image), CHRONODELTA (estimation of time elapsed between two images on a logarithmic scale), CHRONOREVERSE (detection of temporally reversed sequences), and CHRONOODD (identification of a temporal outlier within a set). The benchmark comprises 1{,}000 items across eight process families (biological growth, food transformation, physical weathering, construction, environmental change, human ageing, astronomical phenomena, and urban dynamics) spanning timescales from minutes to millennia. We evaluate eight open-source VLMs (500 M to 19 B parameters) under two prompting regimes and collect human performance baselines. Human performance averages 0.89 across tasks; the best open model (Qwen2.5-VL-7B) reaches 0.40 under direct prompting, a gap we term chronological blindness. Lightweight LoRA fine-tuning on 151 examples raises CHRONODELTA accuracy from near-zero to 0.43, transferring zero-shot to related tasks (CHRONOODD: 0.37; CHRONOREVERSE: 0.64)suggesting the bottleneck is partly instruction following rather than visual perception. Benchmark, code, and predictions will be released upon acceptance.