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 (CS.LG) 2026-06-17

On Randomized Algorithms in Online Strategic Classification

arXiv:2602.06257v2 Announce Type: replace Abstract: Online strategic classification studies settings in which agents strategically modify their features to obtain favorable predictions. For example, given a classifier that determines loan approval based on credit scores, applicants may open or close credit cards and bank accounts to obtain a positive prediction. The learning goal is to achieve low mistake or regret bounds despite such behavior. While randomized algorithms have the potential to offer advantages to the learner in strategic settings, they have been largely underexplored. In the realizable setting, no lower bound is known for randomized algorithms, and existing lower bound constructions for deterministic learners can be circumvented by randomization. In the agnostic setting, the best known regret upper bound is $O(T^{3/4}\log^{1/4}T|\mathcal H|)$, which is far from the standard online learning rate of $O(\sqrt{T\log|\mathcal H|})$. In this work, we provide refined bounds for online strategic classification in both settings; our bounds depend on the Littlestone dimension $\mathrm{Ldim}(\mathcal H)$ of the hypothesis class $\mathcal H$ and the maximum degree $\Delta$ of the manipulation graph. In the realizable setting, we extend, for $T > \mathrm{Ldim}(\mathcal H) \Delta^2$, the existing lower bound $\Omega(\mathrm{Ldim}(\mathcal H) \Delta)$ for deterministic learners to all learners. This yields the first lower bound that applies to randomized learners. We then provide the first randomized learner that improves the known (deterministic) upper bound of $O(\mathrm{Ldim}(\mathcal H) \cdot \Delta \log \Delta)$. In the agnostic setting, we give an improper randomized learner that improves the regret upper bound to $O(\sqrt{T\log|\mathcal H|})$, matching the standard online learning rate. We also show a larger lower bound for all proper learning rules, demonstrating that improperness is necessary to achieve the optimal rate.

02.
arXiv (CS.LG) 2026-06-19

BLISS: A Lightweight Bilevel Influence Scoring Method for Data Selection in Language Model Pretraining

arXiv:2510.06048v5 Announce Type: replace Abstract: Effective data selection is essential for pretraining large language models (LLMs), enhancing efficiency and improving generalization to downstream tasks. However, existing approaches often require leveraging external pretrained models, making it difficult to disentangle the effects of data selection from those of the external pretrained models. In addition, they often overlook the long-term impact of selected data if the model is trained to convergence, primarily due to the prohibitive cost of full-scale LLM pretraining. In this paper, we introduce BLISS (BileveL Influence Scoring method for data Selection): a lightweight data selection method that operates entirely from scratch, without relying on any external pretrained oracle models, while explicitly accounting for the long-term impact of selected data. BLISS leverages a small proxy model as a surrogate for the LLM and employs a score model to estimate the long-term influence of training samples if the proxy model is trained to convergence. We formulate data selection as a bilevel optimization problem, where the upper-level objective optimizes the score model to assign importance weights to training samples, ensuring that minimizing the lower-level objective (i.e., training the proxy model over the weighted training loss until convergence) leads to best validation performance. Once optimized, the trained score model predicts influence scores for the dataset, enabling efficient selection of high-quality samples for LLM pretraining. We validate BLISS by pretraining 410M/1B/2.8B Pythia and LLaMA-0.5B models on selected subsets of the C4 dataset. Notably, under the 1B model setting, BLISS achieves $1.7\times$ speedup in reaching the same performance as the state-of-the-art method, demonstrating superior performance across multiple downstream tasks.

03.
arXiv (CS.LG) 2026-06-19

Score Approximation for Diffusion Models on Arbitrary Low-Dimensional Structures

arXiv:2606.19894v1 Announce Type: new Abstract: The remarkable success of score-based diffusion models has spurred significant efforts to establish their theoretical foundations. However, existing complexity bounds for score approximation rely heavily on restrictive assumptions like Lipschitz continuous densities or smooth manifold supports, which are routinely violated by the singularities, sharp boundaries, and disjoint clusters inherent to real-world perceptual data. This work establishes a universal score approximation theorem that works for any distribution supported on any compact set of upper Minkowski dimension $d$. Using a novel discrete-mixture formulation, we prove that the score function can be approximated with a ReLU network whose complexity grows exponentially only with $d$, thus breaking the exponential curse of ambient dimensionality. Combined with existing theories on accurately solving the backward diffusion SDE for arbitrary compact distributions, our work shows that diffusion models readily adapt to irregular, non-smooth data structures, explaining their competence in real-world generative tasks.

04.
medRxiv (Medicine) 2026-06-11

Global population frequencies of NAT2 star alleles observed in three large biobanks

NAT2 is an important pharmacogene which encodes the N-acetyltransferase 2 enzyme that is involved in the metabolism of multiple medications, and variants in this gene can affect patient response to these medications. CPIC has published a clinical guideline for prescribing hydralazine using NAT2 genotypes. Just prior to the guideline, updated NAT2 star allele numbering and definitions were released, differing somewhat from the historical nomenclature. Clinical pharmacogenomic testing panels often test for the most common star alleles, so knowledge of the most common updated NAT2 star alleles is critical for the implementation of the CPIC NAT2/hydralazine guideline. We first determine NAT2 diplotype frequencies from UK Biobank (UKBB) 200k phased genomes, then analyzed allele, diplotype, and phenotype population frequencies from the All of Us Research program, PennMedicine BioBank (PMBB) and UKBB 500k datasets. We found that analyzing NAT2 diplotypes from phased data provides critical information for algorithms designed to predict diplotypes from unphased data. We observed that NAT2*5, *6, and *4 were the most common star alleles in that order, and the top 11 most frequent NAT2 star alleles were the same across all biobanks. However, differences in star allele frequencies across biogeographical populations were observed. The largest difference led to a higher frequency of NAT2 poor metabolizer phenotypes as compared to rapid and intermediate metabolizer phenotypes in all global populations except in the EAS population, where NAT2 poor metabolizers were in the minority.

05.
medRxiv (Medicine) 2026-06-12

Association of circulating endothelial progenitor cell count and functional outcome in patients with acute ischemic stroke due to intracranial large vessel occlusion

Background: Circulating endothelial progenitor cells (cEPCs) contribute to vascular repair following an ischemic stroke. The aim of the study was to evaluate the association between cEPCs and functional outcomes in patients with acute ischemic stroke (AIS) due to large vessel occlusion (LVO) who received endovascular therapy (EVT). Methods: Prospective study of patients with LVO-AIS who received EVT. Blood samples were obtained within 24 +- 12 hours and on day 7+-1 from stroke onset. cEPCs were detected using flow cytometry (CD34+/VEGFR2+/CD133+). The primary endpoint was a favourable functional outcome (modified Rankin Scale 0-2) at three months of follow-up. Secondary endpoints include baseline to 24 hours/day 7 changes in the National Institutes of Health Stroke Scale (NIHSS) score and collateral circulation (CC) status. Bivariate and multivariable logistic regression analyses were performed. Results: Included were 90 patients (73.2+-12.7 years, 41.1% women) in 42 of whom (46.7%) cEPCs were detected at 24 hours. On day 7, cEPCs were detected in 27 (43.6%) of 62 patients for which this information was available. Atrial fibrillation, prior anticoagulant treatment and stroke onset-to-door time

06.
medRxiv (Medicine) 2026-06-15

Nocturnal Respiratory Rate and Variability Predict Long-term Mortality in Stable Outpatients with Cardiovascular Disease

Background: Respiratory rate (RR) predicts short-term mortality in acute care settings, yet its prognostic significance in clinically stable outpatients remains poorly defined. Objectives: To determine whether the median and variability of nocturnal respiratory rate (NRR) are independently associated with long-term cardiovascular and all-cause mortality in outpatients with cardiovascular disease. Methods: We analyzed overnight chest belt waveforms from elective polysomnography in 5,679 older adults with cardiovascular disease enrolled in the Sleep Heart Health Study (SHHS). NRR was quantified at 30-second resolution, and per-subject median NRR and within-night variability (standard deviation) were derived. Kaplan-Meier survival analysis and Cox proportional hazards models were used to evaluate associations with cardiovascular and all-cause mortality over 3-year and 15-year follow-up periods, adjusting for demographic characteristics, cardiopulmonary comorbidities, and sleep apnea severity. Results: Higher median NRR and greater NRR variability were each associated with increased cardiovascular and all-cause mortality. Combining these metrics identified a high-risk group characterized by elevated median and high variability of NRR, with approximately five-fold higher 3-year all-cause mortality compared with a low-risk group; this association remained significant in Cox models (unadjusted HR: 2.61; 95% CI: 1.65, 4.14; p

07.
arXiv (CS.CV) 2026-06-16

Metis: A Generalizable and Efficient World-Action Model for Autonomous Driving and Urban Navigation

World action models~(WAMs) have shown great promise for autonomous driving and urban navigation. Built upon Vision-Language-Action models or video generation models, existing approaches suffer key limitations: (1) High inference latency due to future observation prediction at test time, and (2) tightly coupled video and action modeling leading to representational mismatch and degraded generalization. To address both issues, we propose Metis, an end-to-end WAM framework that decouples video generation and action prediction. Specifically, Metis employs a Mixture-of-Transformers architecture with dedicated experts for video generation and action prediction, preserving the intrinsic distributional properties of each task. To enhance efficiency, we introduce an asymmetric attention mask that enables joint training of both experts while allowing the action model to bypass explicit video generation during inference. This design ensures training-inference consistency and significantly reduces computational costs without compromising planning performance. Extensive experiments demonstrate state-of-the-art performance on the NAVSIM navhard and navtest benchmarks and the CityWalker navigation benchmark, validating both the generalizability and efficiency across diverse tasks. Real-robot deployments further confirm the practical feasibility of our approach.

08.
arXiv (CS.CV) 2026-06-16

Wasserstein Equilibrium Decoding for Reliable Medical Visual Question Answering

Small vision-language models (2-8B) are well-suited for clinical deployment due to privacy constraints, limited connectivity, and low-latency requirements favouring on-device or on-premise inference. However, their limited capacity exacerbates the generation of plausible but incorrect outputs. We extend game-theoretic decoding, previously restricted to text-only, closed-ended NLP tasks, to vision-language models for open-ended Medical VQA. We introduce a semantically aware Wasserstein stopping criterion that replaces lexical order matching, enabling convergence based on semantic consensus among near-synonymous candidate answers and avoiding unnecessary iterations caused by clinically equivalent ranking swaps. On VQA-RAD and PathVQA, we obtain consistent, statistically significant improvements over greedy and discriminative baselines. On VQA-RAD, we improve Qwen3-VL-2B by +3.5 percentage points (p < 0.01), surpassing the greedy 4B model, with similar trends at larger scales. On PathVQA, Gemma-3-4B with BDG matches MedGemma-4B under greedy decoding despite no domain-specific fine-tuning. At accuracy parity with classic BDG, the Wasserstein criterion reduces average convergence iterations by approximately 20%, improving inference efficiency while preserving the game-theoretic equilibrium behaviour. Code is available at https://github.com/luca-hagen/ Wasserstein-BDG-medical-VQA.

09.
Nature (Science) 2026-06-17

Probing picometre-scale interlayer deformations via hyperbolic polaritons

作者:

The resilience of van der Waals (vdW) materials to large strain fields makes them an ideal platform for tuning electronic, optical and magnetic properties1–4. Although in-plane strain is readily mapped, non-invasive and quantitative characterization of out-of-plane strain remains a formidable challenge, particularly for picometre-scale deformations buried at interfaces. Here we demonstrate a polaritonic optical method that uses the mid-infrared out-of-plane hyperbolic polaritons (oHPs) mode to detect interlayer deformations in prototypical vdW polar insulator–hexagonal boron nitride (hBN). This method uses the softening mechanism of out-of-plane transverse optical (oTO) phonons induced by interlayer strain, enabling highly sensitive detection of picometre-scale deformations. Although these oTO phonon modes are typically spectroscopically ‘dark’, their strain response is activated through the oHPs, achieving an atomic displacement sensitivity of about 10 pm (about 8 × 10−7 times the probing wavelength), enabling ultradeep-subwavelength mechanical interlayer deformation detection. This is experimentally validated in both planar hBN and at the buried interface of quantum dot–hBN nanotube heterostructures. This polariton-based picometrology bridges nanomechanics and photonics, providing a non-destructive lens to visualize hidden stress landscapes with atomic precision. A new polaritonic optical method that uses the mid-infrared out-of-plane hyperbolic polaritons mode is described and experimentally validated to allow the examination of picometre-scale interlayer deformations, providing a bridge between nanomechanics and photonics.

10.
arXiv (CS.AI) 2026-06-18

Robust Regularized Policy Iteration under Transition Uncertainty

arXiv:2603.09344v3 Announce Type: replace Abstract: Offline reinforcement learning (RL) enables data-efficient and safe policy learning without online exploration, but its performance often degrades under distribution shift. The learned policy may visit out-of-distribution state-action pairs where value estimates and learned dynamics are unreliable. To address policy-induced extrapolation and transition uncertainty in a unified framework, we formulate offline RL as robust policy optimization, treating the transition kernel as a decision variable within an uncertainty set and optimizing the policy against the worst-case dynamics. We propose Robust Regularized Policy Iteration (RRPI), which replaces the intractable max-min bilevel objective with a tractable KL-regularized surrogate and derives an efficient policy iteration procedure based on a robust regularized Bellman operator. We provide theoretical guarantees by showing that the proposed operator is a $\gamma$-contraction and that iteratively updating the surrogate yields monotonic improvement of the original robust objective with convergence. Experiments on D4RL benchmarks demonstrate that RRPI achieves strong average performance, outperforming recent baselines including percentile-based methods on the majority of environments while remaining competitive on the rest. Moreover, RRPI exhibits robust performance by aligning lower $Q$-values with high epistemic uncertainty, which prevents the policy from executing unreliable out-of-distribution actions.

11.
arXiv (CS.CL) 2026-06-15

LLMs Contain Multitudes: How Deployment Context Reshapes Model-Level Preferences and Values

Large language models (LLMs) are increasingly characterised in recent evaluation work as having stable, model-level preference and value systems. However, accompanying robustness checks are limited to incidental prompt perturbations such as syntax variation and option reordering. This leaves open whether the measured properties survive when the surrounding task context changes, as it does in most real deployments. We test this directly across two established pairwise paradigms: ranking country preferences and eliciting utility judgements. In both, we make the deployment context – the high-level task the model is performing while making concrete value-dependent choices – our controlled variable, varied across framings such as writing a Reddit post or a news article. Across five LLMs and over 1.2M pairwise decisions, deployment context produces variation far larger than prompt paraphrasing and temperature controls. In country preference rankings over 15 countries, context induces widespread, statistically significant rank shifts; the aggregate Global North favouritism reported in prior work is itself context-dependent, with each model's bias shifting systematically across contexts. In utility elicitation over 50 outcomes, broad cross-category ordering is preserved, but fine-grained rankings within domains vary substantially, and cardinal exchange rates between outcomes (e.g. how many lives in one region equal one in another) shift by a factor of 2.47 at the median. Reported model-level preferences and utilities are therefore better understood as context-conditioned measurements than fixed model-level properties: safety guarantees obtained under one framing provide limited assurance in another.

12.
arXiv (CS.LG) 2026-06-19

Beyond Averaging in John Ellipsoid Approximation: High-Accuracy Algorithms in the Leverage-Score Model

arXiv:2606.20082v1 Announce Type: cross Abstract: The John ellipsoid of a symmetric polytope $P=\{\mathbf{x}\in\mathbb{R}^d:\|\mathbf{A}\mathbf{x}\|_\infty\le1\}$, $\mathbf{A}\in\mathbb{R}^{n\times d}$, is computed by a long line of leverage-score algorithms, from Cohen, Cousins, Lee and Yang (COLT 2019) to its successors [WY24, CLS+25], all reaching a $(1+\varepsilon)$-approximation in $\Theta(\varepsilon^{-1}\log(n/d))$ iterations. We separate this complexity into three costs the modern line conflates (certification, identification, and accuracy) and locate the historical $\varepsilon^{-1}$ in the first alone. In the equivalent D-optimal-design form $\min_{\mathbf{p}\in\Delta_n}-\log\det(\sum_i p_i\mathbf{a}_i\mathbf{a}_i^\top)$, the leverage-score oracle is exactly the first-order oracle and the $(1+\varepsilon)$-John guarantee the Frank-Wolfe gap $g(\mathbf{p})\le\varepsilon d$; through this dictionary the costs come apart. The $\varepsilon^{-1}$ is a certification artifact: the uniform average of the iterates, the certificate used throughout the line, has gap exactly $\Theta(1/T)$, however cheap each iteration is made. Pointed instead at the last iterate the same oracle is fast: a warm-started accelerated method reaches the guarantee in $C(\mathbf{A})+O(\sqrt{\kappa}\log(1/\varepsilon))$ queries after an $\varepsilon$-independent setup $C(\mathbf{A})$, and once the optimal face is identified the facial problem is an unconstrained self-concordant minimization whose Hessian the oracle recovers exactly, so damped Newton needs only $O(\log\log(1/\varepsilon))$ steps, for a total of $C(\mathbf{A})+O(d^2\log\log(1/\varepsilon))$ queries. The accuracy dependence is thus doubly logarithmic after an $\varepsilon$-independent, condition-dependent setup; the open problem is the remaining identification cost (a condition-free bound on reaching the optimal face) and lower bounds. Accuracy is not the obstruction.

13.
arXiv (CS.LG) 2026-06-15

Deep Doubly Debiased Longitudinal Effect Estimation with ICE G-Computation

arXiv:2602.12379v2 Announce Type: replace Abstract: Estimating longitudinal treatment effects is essential for sequential decision-making but is challenging due to treatment-confounder feedback. While Iterative Conditional Expectation (ICE) G-computation offers a principled approach, its recursive structure suffers from error propagation, corrupting the learned outcome regression models. We propose D3-Net, a framework that mitigates error propagation in ICE training and then applies a robust final correction. First, to interrupt error propagation during learning, we train the ICE sequence using Sequential Doubly Robust (SDR) pseudo-outcomes, which provide bias-corrected targets for each regression. Second, we employ a multi-task transformer with a covariate simulator head for auxiliary supervision, regularizing representation learning, and a target network to stabilize training dynamics. For the final estimate, we discard the SDR correction and instead use the uncorrected nuisance models to perform Longitudinal Targeted Minimum Loss-Based Estimation (LTMLE) on the original outcomes. This second-stage, targeted debiasing ensures robustness and optimal finite-sample properties. Comprehensive experiments demonstrate that our model, D3-Net, robustly reduces bias and variance across different horizons, counterfactuals, and time-varying confoundings, compared to existing state-of-the-art ICE-based estimators.

14.
arXiv (math.PR) 2026-06-17

Decay of correlations and zeros for the hard-core model

arXiv:2603.17858v2 Announce Type: replace Abstract: In a recent paper the last author proved that absence of complex zeros of the partition function of the hard-core model near a parameter $\lambda>0$ implies a form of correlation decay called strong spacial mixing. In this paper we investigate the reverse implication. We introduce a strengthening of strong spatial mixing that we call very strong spatial mixing (VSSM). Our main result is that if VSSM holds at a parameter $\lambda>0$ for a family of graphs, this implies that the partition function has no zeros near that parameter for each graph in the family. We also demonstrate that a closely related variant of very strong spatial mixing does not imply zero-freeness. As a consequence of our main result, we moreover obtain that VSSM implies spectral independence. Our proof relies on transforming the problem to the analysis of an induced non-autonomous dynamical system given by Möbius transformations.

15.
arXiv (CS.CV) 2026-06-18

BindEdit: Taming Attention Leakage for Precise Multi-Object Image Editing

Real image editing enables precise manipulation of visual content, yet existing methods often fail in complex multi-object scenarios, causing semantic blending, object duplication, or incomplete edits. We attribute these failures to attention leakage, where signals across spatial regions and text tokens become entangled during the denoising process. Specifically, we identify two distinct forms of leakage: Edit-Token Leakage, where ambiguous token-region alignment leads to object blending, and Source Dominance Leakage, where tokens of unchanged source objects overwhelm the attention intended for target entities. To resolve these leakages, we propose BindEdit, which enforces attention-level constraints within a single diffusion trajectory. To suppress Edit-Token Leakage, BindEdit jointly regularizes cross- and self-attention so that each target token group is bound to its corresponding spatial region while maintaining instance-level separation. To suppress Source Dominance Leakage, a cross-attention re-balancing mechanism amplifies target token influence and attenuates residual source semantics within editable regions. Moreover, a region fidelity term ensures that each target concept is expressed coherently across the entire editing mask. Additionally, we propose a comprehensive multi-object benchmark encompassing diverse object counts and categories. Extensive experiments demonstrate that BindEdit consistently outperforms existing methods within a single diffusion trajectory, maintaining robust performance across both single- and multi-object editing scenarios.

16.
arXiv (CS.CV) 2026-06-11

Slots, Transitions, Loops: Learning Composable World Models for ARC

ARC tests in-context rule induction: given a few input-output demonstrations, a model must infer the hidden rule and apply it to a new query. While many approaches express ARC rules through language, code, or symbolic programs, ARC itself is visual-symbolic: rules appear as grid transitions over objects, colors, shapes, and spatial relations. We introduce Loop-OWM, an object-centric world-modeling architecture that learns these rules as composable transitions over structured states. It combines color-prototype slots, demonstration-conditioned task summaries, and a looped transition model with dense propagation and slot-conditioned correction. On both ARC-1 and ARC-2, Loop-OWM outperforms non-looped and looped baselines with comparable or fewer parameters. These results suggest that ARC rules can be learned not only as language descriptions or searched programs, but also as transitions over visual-symbolic world states.

17.
arXiv (CS.CL) 2026-06-18

Which Sections of a Research Paper Best Reveal Its Research Methods? Evidence from Library and Information Science

Research methods are essential carriers of knowledge contribution in academic papers. Automatic multi-label classification of research methods can support knowledge services such as method retrieval, review generation, and research intelligence analysis. While existing studies primarily rely on titles and abstracts, abstracts often provide only limited methodological information, whereas utilizing full-text content faces challenges related to excessive length and information redundancy. Therefore, this paper proposes a segment combination strategy by partitioning the full-text content according to its physical postion. Using an annotated corpus of 1,954 full-text articles from three representative journals in Library and Information Science (JASIST, LISR, and JDoc), we evaluate the classification performance of various segments and their combinations across multiple models. Experimental results indicate that methodological information is distributed unevenly within the full-text content, with the middle-to-late and final segments exhibiting greater discriminative power. Furthermore, integrating bibliographic metadata with cross-segment combination strategies effectively enhances classification performance.

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

XFlow: An Executable Protocol Programming System for Reliable Multi-Agent Workflows

arXiv:2606.14790v1 Announce Type: cross Abstract: LLM-based multi-agent systems increasingly coordinate planning, reasoning, tool use, and human interaction, yet their reliability remains limited. A central source of this limitation is the underspecified prompt–harness boundary. Current systems lack a principled way to decide which workflow commitments should remain in prompts and which should become harness structure. We present XFlow, an executable protocol programming system for reliable multi-agent workflows, and XPF (XFlow Protocol Format), its domain-specific protocol programming language. XFlow occupies a middle position between prompt-only orchestration and markup-like workflow descriptions. XPF remains readable as a literate protocol, but it is compiled and executed as a program. Its design keeps informal semantic work inside actors while moving selected commitments into harness structure that can be checked, preserved, and enforced. At runtime, XFlow stages uncertainty through lifecycle-governed symbols, which are typed state cells with validation and commit states. Actor outputs are mediated before they become shared state, instead of spreading through prompts, transcripts, or implicit memory. Our experiments cover Constrained Interaction, Long-Context Reasoning, and Agentic Software Engineering. They show that XFlow improves reliability by making constraints, evidence handling, and process requirements explicit and enforceable.

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

When Probing Accuracy Saturates, Fragility Resolves: A Complementary Metric for LLM Pre-Training Analysis

Standard linear probing declares a property "encoded" when a classifier on hidden states achieves high accuracy. The protocol works well on a snapshot but breaks across pre-training: probe accuracy saturates within the first few thousand steps, leaving most of training invisible to the instrument. We introduce fragility, a complementary per-layer metric defined as the activation-noise level at which probe accuracy collapses. Fragility is sensitive to both the margin of separability and the redundancy of representation, both of which keep evolving long after accuracy plateaus. Applied to open-checkpoint language models, fragility recovers structure that accuracy alone cannot see. Moralized representations emerge along a lexical $\to$ compositional gradient: lexical moral detection first, compositional moral encoding later. Because probe accuracy on its own tracks how lexically separable a dataset is, we establish the compositional encoding directly, by showing it transfers across construction types that share no contrast tokens. A layer-depth robustness gradient develops monotonically across training while accuracy stays flat. And matched fine-tuning corpora that produce identical probing accuracy leave distinct fragility fingerprints, showing that data curation reshapes probe robustness without changing probe accuracy. In every comparison we test, where probing accuracy returns a flat answer, fragility returns a structured one.

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

Efficient Rationale-based Retrieval: On-policy Distillation from Generative Rerankers based on JEPA

Unlike traditional fact-based retrieval, rationale-based retrieval typically necessitates cross-encoding of query-document pairs using large language models, incurring substantial computational costs. To address this limitation, we propose Rabtriever, which independently encodes queries and documents, while providing comparable cross query-document comprehension capabilities to rerankers. We start from training a LLM-based generative reranker, which puts the document prior to the query and prompts the LLM to generate the relevance score by log probabilities. We then employ it as the teacher of an on-policy distillation framework, with Rabtriever as the student to reconstruct the teacher's contextual-aware query embedding. To achieve this effect, Rabtriever is first initialized from the teacher, with parameters frozen. The Joint-Embedding Predictive Architecture (JEPA) paradigm is then adopted, which integrates a lightweight, trainable predictor between LLM layers and heads, projecting the query embedding into a new hidden space, with the document embedding as the latent vector. JEPA then minimizes the distribution difference between this projected embedding and the teacher embedding. To strengthen the sampling efficiency of on-policy distillation, we also add an auxiliary loss on the reverse KL of LLM logits, to reshape the student's logit distribution. Rabtriever optimizes the teacher's quadratic complexity on the document length to linear, verified both theoretically and empirically. Experiments show that Rabtriever outperforms different retriever baselines across diverse rationale-based tasks, including empathetic conversations and robotic manipulations, with minor accuracy degradation from the reranker. Rabtriever also generalizes well on traditional retrieval benchmarks such as MS MARCO and BEIR, with comparable performance to the best retriever baseline.

21.
arXiv (CS.AI) 2026-06-19

ParaScale: Scale-Calibrated Camera-Motion Transfer via a Gauge-Invariant Parallax Number

作者:

arXiv:2606.19805v1 Announce Type: cross Abstract: Transferring the camera motion of a reference video to a freshly generated one lets creators reuse cinematic moves. Yet reference and target often live at incompatible scales – a sweep across a galaxy versus a nudge across a desk – and naively reusing the recovered trajectory yields either imperceptible or violently exaggerated motion. We trace this to a geometric fact: translation-induced image motion scales as ||T||/Z, so a monocular trajectory is meaningful only up to a depth-scale gauge. We distill this into the Parallax Number Pi = ||Delta T|| / Zbar, a dimensionless, gauge-invariant descriptor of how strongly a camera move is felt, and prove that it – not the raw trajectory – is the quantity that scale-faithful transfer must preserve. ParaScale is a plug-and-play module that reads Pi off any reference video and re-realizes it against the target scene's own depth, per frame, leaving rotation untouched. Sitting between pose extraction and pose injection, it requires no retraining and drops into any pose-conditioned generator. We further introduce the Parallax Consistency Error (PCE), a scale-symmetric metric that – unlike the similarity-aligned TransErr – exposes scene-scale mismatch. Across scale regimes spanning four orders of magnitude and multiple backbones, ParaScale keeps the realized parallax on the identity line and cuts PCE by more than 3x over uncalibrated transfer with no loss of visual fidelity.

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

Unsafer in Many Turns: Benchmarking and Defending Multi-Turn Safety Risks in Tool-Using Agents

LLM-based agents are becoming increasingly capable, yet their safety lags behind. This creates a gap between what agents can do and should do. This gap widens as agents engage in multi-turn interactions and employ diverse tools, introducing new risks overlooked by existing benchmarks. To systematically scale safety testing into multi-turn, tool-realistic settings, we propose a principled taxonomy that transforms single-turn harmful tasks into multi-turn attack sequences. Using this taxonomy, we construct MT-AgentRisk (Multi-Turn Agent Risk Benchmark), the first benchmark to evaluate multi-turn tool-using agent safety. Our experiments reveal substantial safety degradation: the Attack Success Rate (ASR) increases by 16% on average across open and closed models in multi-turn settings. To close this gap, we propose ToolShield, a training-free, tool-agnostic, self-exploration defense: when encountering a new tool, the agent autonomously generates test cases, executes them to observe downstream effects, and distills safety experiences for deployment. Experiments show that ToolShield effectively reduces ASR by 30% on average in multi-turn interactions. Our code is available at https://github.com/CHATS-lab/ToolShield.

23.
arXiv (CS.LG) 2026-06-12

$\mu$VLA: On Recurrent Memory for Partially Observable Manipulation in VLA Models

arXiv:2606.12497v1 Announce Type: new Abstract: Vision-language-action (VLA) models predict chunks of future actions from the current observation, an assumption that fails under partial observability, where decisions depend on information no longer visible. Existing memory-augmented VLAs simultaneously introduce recurrence, retrieval, compression modules, auxiliary objectives, hierarchical memory, or task-specific architectural changes, so the contribution of recurrence itself remains entangled with surrounding machinery. We present a controlled isolation study of recurrence in a strong pretrained VLA backbone. Our formulation augments the transformer with a small set of learnable memory tokens carried across timesteps and updated through self-attention, trained end to end with truncated backpropagation through time, with no auxiliary losses and no architectural changes. We instantiate this as $\mu$VLA, a family of OpenVLA-OFT variants parameterized by memory width m, TBPTT length K, and the memory update rule (cross-step gradients or a detached EMA), so that recurrence is the only varying factor. On MIKASA-Robo, $\mu$VLA improves average success rate on five training tasks from 0.42 to 0.84 at the strongest setting and reaches 0.23 on held-out tasks with the same memory structure versus 0.07 for the memoryless baseline. On tasks requiring different memory structure, performance remains near baseline. On LIBERO, the strongest recurrent variant achieves 96.2% average success, indicating no regression under full observability. We interpret these results as a calibration of the capability envelope of minimal in-backbone recurrence, identifying the regime in which it is sufficient and the regime where additional memory structure is required. Demos and videos can be found in https://avanturist322.github.io/mu-vla/.

24.
arXiv (CS.AI) 2026-06-16

Adaptive Memory Crystallization for Autonomous AI Agent Learning in Dynamic Environments

arXiv:2604.13085v2 Announce Type: replace-cross Abstract: Autonomous AI agents operating in dynamic environments face a persistent challenge: acquiring new capabilities without erasing prior knowledge. We present Adaptive Memory Crystallization (AMC), a memory architecture for progressive experience consolidation in continual reinforcement learning. AMC is conceptually inspired by the qualitative structure of synaptic tagging and capture (STC) theory, the idea that memories transition through discrete stability phases, but makes no claim to model the underlying molecular or synaptic mechanisms. AMC models memory as a continuous crystallization process in which experiences migrate from plastic to stable states according to a multi-objective utility signal. The framework introduces a three-phase memory hierarchy (Liquid–Glass–Crystal) governed by an Itô stochastic differential equation (SDE) whose population-level behavior is captured by an explicit Fokker–Planck equation admitting a closed-form Beta stationary distribution. We provide proofs of: (i) well-posedness and global convergence of the crystallization SDE to a unique Beta stationary distribution; (ii) exponential convergence of individual crystallization states to their fixed points, with explicit rates and variance bounds; and (iii) end-to-end Q-learning error bounds and matching memory-capacity lower bounds that link SDE parameters directly to agent performance. Empirical evaluation on Meta-World MT50, Atari 20-game sequential learning, and MuJoCo continual locomotion consistently shows improvements in forward transfer (+34–43\% over the strongest baseline), reductions in catastrophic forgetting (67–80\%), and a 62\% decrease in memory footprint.

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

DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning

Reasoning capabilities of multimodal large language models (MLLMs) have improved considerably in recent years. Existing approaches typically rely on explicit chain-of-thought or continuous latent-space trajectories to enhance multi-step reasoning. However, these methods generally assume that an input admits a single latent interpretation and unfold reasoning along a fixed path or under a uniform computation budget. In real-world multimodal settings, visual observations are often subject to occlusion, blur, viewpoint variation, or semantic ambiguity, giving rise to multiple plausible interpretations. A uniform reasoning strategy not only limits the model's ability to explore multiple hypotheses but also incurs high memory usage and rollout cost. We present DLWM (Diverse Latent World Models), a multimodal reasoning framework that combines latent-space reasoning with reinforcement learning. First, we construct a set of diverse latent world hypotheses in continuous latent space, each capturing a different plausible interpretation of the visual input, and unfold latent reasoning independently on each hypothesis. An orthogonality-based diversity regularizer explicitly prevents hypothesis collapse. Second, we formulate the latent reasoning process as a resource-constrained sequential decision problem and introduce a resource-aware reinforcement learning policy that adaptively allocates computation across hypotheses, dynamically deciding whether to expand, terminate, or merge reasoning paths, thereby substantially reducing memory footprint and improving rollout efficiency. Experiments on multiple multimodal reasoning benchmarks demonstrate that DLWM outperforms existing methods by 2-5 points in accuracy while reducing memory usage by 24%.