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.CL) 2026-06-16

A Unified Definition of Hallucination: It's The World Model, Stupid!

Despite numerous attempts at mitigation since the inception of language models, hallucinations remain a persistent problem even in today's frontier LLMs. Why is this? We review existing definitions of hallucination and fold them into a single, unified definition wherein prior definitions are subsumed. We argue that hallucination can be unified by defining it as simply inaccurate (internal) world modeling, in a form where it is observable to the user. For example, stating a fact which contradicts a knowledge base OR producing a summary which contradicts the source. By varying the reference world model and conflict policy, our framework unifies prior definitions. We argue that this unified view is useful because it forces evaluations to clarify their assumed reference "world", distinguishes true hallucinations from planning or reward errors, and provides a common language for comparison across benchmarks and discussion of mitigation strategies. Building on this definition, we also connect our framework to HalluWorld, a complementary benchmark that instantiates fully specified reference world models for stress-testing model hallucinations.

02.
arXiv (CS.CV) 2026-06-11

Natural-Language Temporal Grounding in Hour-Long Videos is a Search Problem: A Benchmark and Empirical Decomposition

Temporal grounding–returning the interval $[t_s, t_e]$ for a natural-language query over a video–is the language interface to long-form video, yet has been studied on short videos; the dynamics of hour-scale natural-language grounding remain underexplored. We take the position that at hour-scale, the binding constraint is search, not recognition: Video-LLMs are bottlenecked not by localizing a nearby event, but–given a natural-language query–by searching for the relevant region of a long video. To test this, we release ExtremeWhenBench, the first open hour-scale grounding benchmark (2,273 queries over 194 videos, mean 75.7 min, max 9 hr) with an open-form query distribution. Every open Video-LLM collapses while a frame-level retrieval baseline outperforms them; a failure taxonomy attributes 85% of failures to search; and a retrieve-then-ground hybrid recovers 6.7x over the monolithic Video-LLM–mirroring retrieve-then-read in open-domain QA.

03.
arXiv (CS.AI) 2026-06-17

LLM-Powered Multi-Agent System for Automated Crypto Portfolio Management

arXiv:2501.00826v3 Announce Type: replace-cross Abstract: Cryptocurrency portfolio management requires the fusion of heterogeneous multi-modal signals, including structured price and on-chain time series, unstructured news text, and technical indicators, under high-volatility and real-time constraints. While deep learning approaches show predictive capability, their opacity limits practical adoption, and single large language model (LLM) agents struggle to process the breadth of modality-specific inputs needed for robust decision-making. We propose a multi-agent system (MAS) framework in which three modality-specialised agents, a Crypto Agent for market dynamics, a News Agent for weekly news sentiment, and a Trading Agent for signal fusion and portfolio execution, decompose the task across three communication architectures: hierarchical, collaborative, and debate. We evaluate four capability configurations: zero-shot, chain-of-thought (CoT), retrieval-augmented generation (RAG), and skill-augmented. In a 52-week backtest over calendar year 2025 across the top 15 L1 blockchain native cryptocurrencies by market capitalisation as of January 2025, the best configuration, Hierarchical (Skill), achieves a cumulative return of 133.52% and a Sharpe ratio of 1.502, outperforming single-agent variants, passive benchmarks, and deep learning baselines. An ablation study identifies the Crypto Agent as the most critical component, with its removal reducing cumulative return by 42.57 percentage points. A cross-model comparison further shows that MAS outperforms the single-agent baseline under GPT-4o, GPT-5, and Claude Sonnet 4.5, suggesting that the benefit of multi-agent coordination is model-agnostic. Unlike black-box deep learning models, every portfolio decision is traceable to explicit agent reasoning, offering an interpretable and effective approach to multi-modal cryptocurrency portfolio management.

04.
arXiv (CS.LG) 2026-06-18

CODEBLOCK: Learning to Supervise Code at the Right Granularity

arXiv:2606.18286v1 Announce Type: new Abstract: Supervised fine-tuning of code LLMs typically applies uniform cross-entropy loss to all response tokens, implicitly assuming that every token provides equally useful learning signal. Recent token-level selection methods challenge this assumption in natural-language SFT by supervising only high-value tokens. However, directly transferring token-level masking to code can break syntactically and semantically coherent program units, because code depends on structural completeness and definition-use relations. We therefore propose CodeBlock, a structure-aware sparse supervision framework that selects structure-complete code evidence rather than isolated tokens. CodeBlock first selects high-quality instruction-response pairs, then partitions code responses into syntactically coherent coding items, estimates their utility by aggregating generalized cross-entropy over core logic tokens, and reranks them with data-flow reach and bridge signals to prioritize blocks that propagate or connect important program dependencies. During training, the full response remains available as context, while loss is applied only to selected code items and informative natural-language tokens. Experiments on six code-generation benchmarks show that CodeBlock achieves stronger average pass@1 than full-token SFT and competitive selection baselines, while using only 1.9% of supervised response tokens.

05.
arXiv (CS.CV) 2026-06-15

Digital Twin Driven Textile Classification and Foreign Object Recognition in Automated Sorting Systems

The increasing demand for sustainable textile recycling requires robust automation solutions capable of handling deformable garments and detecting foreign objects in cluttered environments. This work presents a digital twin driven robotic sorting system that integrates grasp prediction, multi modal perception, and semantic reasoning for real world textile classification. A dual arm robotic cell equipped with RGBD sensing, capacitive tactile feedback, and collision-aware motion planning autonomously separates garments from an unsorted basket, transfers them to an inspection zone, and classifies them using state of the art Visual Language Models (VLMs). We benchmark nine VLM s from five model families on a dataset of 223 inspection scenarios comprising shirts, socks, trousers, underwear, foreign objects (including garments outside of the aforementioned classes), and empty scenes. The evaluation assesses per class accuracy, hallucination behavior, and computational performance under practical hardware constraints. Results show that the Qwen model family achieves the highest overall accuracy (up to 87.9 %), with strong foreign object detection performance, while lighter models such as Gemma3 offer competitive speed accuracy trade offs for edge deployment. A digital twin combined with MoveIt enables collision aware path planning and integrates segmented 3D point clouds of inspected garments into the virtual environment for improved manipulation reliability. The presented system demonstrates the feasibility of combining semantic VLM reasoning with conventional grasp detection and digital twin technology for scalable, autonomous textile sorting in realistic industrial settings.

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

PVminerLLM2: Improving Structured Extraction of Patient Voice via Preference Optimization

Motivation: Patient-generated text contains critical information on patients' lived experiences, social context, and care engagement, but remains largely unstructured, limiting its use in patient-centered outcomes research. Prior work introduced the PV-Miner benchmark and PVMinerLLM models for structured extraction. However, supervised fine-tuning (SFT) alone struggles with rare, fine-grained, and unevenly distributed errors, particularly in token-critical structured outputs. Results: We present PVminerLLM2, an improved set of LLMs for structured patient voice extraction that applies preference optimization to address token-critical errors beyond the reach of supervised fine-tuning. Our method introduces (i) a preference objective with token-level gated stabilization term that prevents degradation of absolute token likelihood under preference optimization, and (ii) confusion-aware preference pair construction to better capture low-separation distinctions. We further incorporate token-importance weighting and inverse-frequency reweighing to address token imbalance and class skew. Across multiple model sizes, PVMinerLLM2 consistently outperforms strong baselines, achieving gains of up to 4.43% (Code), 3.50% (Sub-code), and 1.55% (Span), and outperforms baseline LLM trained with existing preference optimization methods. Availability and Implementation: The supplementary material, code, evaluation scripts, and trained models for PVminerLLM2 are publicly available at: https://github.com/Data-Mining-Lab-Yale/PVminerLLM2

07.
arXiv (CS.CL) 2026-06-12

Small LLMs for Biomedical Claim Verification: Cost-Effective Fine-Tuning, Structural Dataset Shortcuts, and Cross-Domain Generalization

作者:

Large Language Models such as GPT-4o and GPT-5 achieve strong zero-shot performance on biomedical claim verification, but cost and opacity limit scalable use. We fine-tune three small LLMs: Phi-3-mini (3.8B), Qwen2.5-3B, and Mistral-7B, via QLoRA on SciFact and HealthVer, providing the first study of QLoRA models against GPT-4o and fine-tuned BioLinkBERT encoders. Mistral-7B QLoRA surpasses both GPT-4o and GPT-5 (up to 12% F1 gain) at a fractional cost using just 1,008 training examples. We conduct extensive in-domain and cross-domain evaluation: models trained on SciFact tested on HealthVer and vice versa, at matched sizes to isolate dataset structure from data quantity. We identify a previously unreported structural artifact in SciFact that inflates in-domain scores, and show through bidirectional out-of-domain evaluation that training on structurally sound data enables robust cross-domain transfer. We plan to release all code and adapter checkpoints.

08.
arXiv (CS.CV) 2026-06-15

Visual Quality Score Assessment of Large White Goods in Remanufacture with Multi-View Deformable-DETR

Remanufacturing large white goods is essential for a circular economy, yet visual quality assessment remains a manual bottleneck for training and pricing. Conventional detection methods require extensive annotation and struggle with small defects in high-resolution multi-view data. We present a multi-view framework based on Deformable-DETR for automated quality scoring that aggregates information across redundant views to extract fine-grained features. To enhance robustness with limited labels, we employ self-supervised pretraining followed by supervised fine-tuning on expert-annotated scores. Additionally, a linear projection over frozen feature maps identifies regions of interest to explain model decisions. Evaluated on an industrial multi-view dataset, our approach delivers precise quality assessments while reducing reliance on manual annotation and per-part customization, enabling scalable and transparent inspection for remanufacturing lines.

09.
medRxiv (Medicine) 2026-06-18

Hard to Halt: Automation Bias in Agent-Driven Sequencing Prior Authorization Workflows

Purpose: Prior authorization (PA) for exome or genome sequencing is a time-consuming process that impedes timely rare disease diagnosis. Large language model-based browser agents offer potential for automating these workflows, but their clinical reliability remain uncharacterized. Methods: We developed a sandbox compromising a simulated ES/GS PA submission payer portal and a synthetic EHR containing 836 patient records spanning compliant profiles and deficient profiles with different types of issues. Gemini 3 Pro, Gemini 3 Flash, and Claude Opus 4.5 were evaluated on task completion rate, form completion accuracy, and appropriate withholding for deficient profiles. Results: Larger models achieved much higher task completion rates (Gemini 3 Pro 95.45%, Claude Opus 4.5 93.67%) compared to Gemini 3 Flash (56.05%), but nearly universally failed to withhold submission for deficient profiles whereas Gemini 3 Flash ironically demonstrated superior withholding performance (17.33%). In a non-agentic setting, Gemini 3 Pro correctly identified 91% of the issues in deficient profiles, indicating that withholding failure is attributable to the browser interaction rather than the model's reasoning limitations. Conclusion: Current LLM-based browser agents exhibit a systematic bias towards form submission that poses risks in PA workflows. A modular, multi-agent architecture with human supervision is necessary for a safe clinical deployment.

10.
PLOS Computational Biology 2026-06-10

Interpreting higher-order dependence in multimorbidity using cohort data: A partial information decomposition approach

by Cillian Hourican, Geeske Peeters, René J. F. Melis, Almar Kok, Natasja M. van Schoor, Sandra Wezeman, Mike Lees, Marcel G. M. Olde Rikkert, Rick Quax In the context of multimorbidity, clinical features seldom act in isolation: symptoms, signs and behaviours form interdependent systems in which joint effects on function can be demonstrated only when features are considered together. We introduce an open, reusable workflow that detects and interprets these “together-only” interactions using bivariate Partial Information Decomposition (PID; two sources to one target), linking synergy-based dependence to the broader network of clinical variables rather than to a single target. The workflow estimates synergy with small-sample bias correction and summarises each pair in a Breadth–Uniformity–Synergy–Total (BUST) map: breadth of synergy across target variables (broad “generalist” vs narrow “specialist” patterns), cross-stratum uniformity across age, sex and multimorbidity (uniform vs subgroup-specific), synergy strength, and total shared information. Simple diagnostics contrast observed targets with additive expectations, revealing the specific joint configurations through which non-additive effects arise. Applied to data from the Longitudinal Ageing Study Amsterdam, we treated all health-related variables—covering symptoms, clinical signs, behaviours, lifestyle factors, and self-rated health indicators—as both sources and targets in the PID framework. This symmetric design permits synergy to be quantified for every pair of variables with respect to every other variable. The workflow identifies synergistic constellations that additive models miss. Multidomain cliques involving subjective health, pain, cognition and grip strength showed multiple non-additive configurations, whereas pairs such as alcohol use with grip strength exhibited focused, narrow but uniform synergy. Notably, the pairs with the strongest synergistic contributions were largely distinct from those with the highest total mutual information, indicating that synergy captures dependency structure overlooked by conventional association measures. Rather than a new measure, this work provides a bias-aware workflow that makes higher-order dependence visible and transferable. Our results support synergy-aware mapping as a practical complement to conventional multimorbidity analyses: it highlights specific combinations of routinely assessed features whose joint states may be especially informative across multiple health targets and therefore candidates for prioritised joint assessment and future multi-domain intervention studies.

11.
PLOS Medicine 2026-05-15

Spatial transcriptomic-metabolic features of tumor foci and tumor capsule in microvascular invasion with hepatocellular carcinoma: A spatial multi-omics study

作者:

by Zhi-Hui Luo, Na Wang, Jingwei Zhao, Fei Long, Si Wu, Wei Zhong, Wei-Ming Chen, Bicheng Wang, Kun Wang, Yufeng Yuan, Jingjiao Zhou, Chunhui Yuan, Fubing Wang Background Microvascular invasion (MVI) is closely related to the recurrence and metastasis of hepatocellular carcinoma (HCC), but the underlying cellular mechanism remains largely elusive. This study aims to elucidate the regional cellular discrepancy between MVI-positive (MVI+) and MVI-negative (MVI−) HCC by integrating Spatial transcriptomics (ST) and spatial metabolomics (SM). Methods and findings ST and SM were performed on six tissue samples from four patients (including 2 MVI+, 2 MVI−, and 2 paratumor tissues), with the integration of 79 public single-cell RNA sequencing datasets of HCC. Patient identity was used as a covariate in the linear equation for regional differentially expressed gene analysis with the ST data. Clinical validation was conducted through multiplex immunofluorescence staining in 79 patients, together with external validation in the cancer genome atlas (TCGA)-liver hepatocellular carcinoma (LIHC) cohort (n = 299) and an independent microarray dataset (n = 62). For cell-type-specific metabolic profiling, spatial transcriptomic-metabolic registration was performed. The functional roles of key metabolites were further validated in vitro using inflammatory cancer-associated fibroblasts (iCAFs) derived from hepatic stellate cells (HSCs) and primary CAFs through co-culture models and various functional assays assessing cell proliferation, migration, and invasion. In the tumor lesion, a malignant STMN1+HMGN2+GPC3+ cell subtype enriched in MVI+ HCC was identified, which exhibited enhanced proliferative activity and was associated with poor prognosis. This finding was further confirmed in a local cohort of 79 patients, where multiplex immunofluorescence staining for the three genes (STMN1, HMGN2, and GPC3) showed significantly higher expression in the MVI+ group than in the MVI− group (p = 0.046). Integrated SM analysis further revealed that this cell population underwent metabolic reprogramming characterized by suppressed glycerolipid metabolism. In the tumor capsule, iCAFs-related genes were downregulated in MVI+ cases, and iCAFs were located distally from the tumor boundary. Spatial metabolite mapping showed a strong correlation between taurine and iCAFs, and functional assays demonstrated that taurine promotes HCC proliferation and migration by suppressing iCAF activity. One limitation of this study is the small sample size of spatial omics data, which hinders a more complete molecular functional analysis of the STMN1+HMGN2+GPC3+ cell subtype and iCAFs in MVI+ HCC. Larger-scale ST cohorts are required to further validate and expand the findings of this study. Conclusions This integrative spatial atlas proposes a hypothesis that there exists a highly proliferative and metabolically reprogrammed malignant cell subtype in the tumor lesion of MVI+ HCC, and that taurine in the tumor capsule modulates iCAF activity to influence tumor progression. The exploratory results provide mechanistic insights into MVI-related HCC progression and offer potential avenues for targeted therapeutic intervention of MVI+ HCC.

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

Beyond Monolingual Deep Research: Evaluating Agents and Retrievers with Cross-Lingual BrowseComp-Plus

Deep research agents are increasingly evaluated on their ability to search for evidence, reason over retrieved sources, and produce grounded answers. Existing browsing benchmarks, however, largely assume that the user's query and the supporting evidence are written in the same language, leaving open whether agentic search systems can operate when relevant evidence appears in another language. We introduce XBCP (Cross-lingual BrowseComp-Plus), a controlled benchmark that preserves the English question-and-answer space of BrowseComp-Plus but varies the languages of the supporting documents. XBCP instantiates two complementary settings: in the cross-lingual setting, each query is paired with evidence in a single assigned language. In the multilingual setting, the full evidence corpus is distributed equally and randomly across 12 languages spanning high-resource and low-resource regimes. We evaluate four deep research agents using sparse and dense multilingual retrievers, measuring answer accuracy, evidence recall, search behavior, calibration, citation fidelity, and oracle retrieval. Results reveal substantial degradation when evidence is translated. Even strong, dense retrievers lose evidence recall, and agents become less calibrated and cite evidence less reliably. Notably, accuracy remains lower even when all gold evidence is supplied directly. These findings suggest that cross-lingual deep research exposes both retrieval failures and an independent, agent-side difficulty in integrating language-mismatched evidence.

13.
arXiv (CS.CL) 2026-06-12

Operads for compositional reasoning in LLMs

Question decomposition, i.e. breaking a complex query into simpler sub-queries whose answers are composed to produce a final answer, is a widely used strategy for improving LLM reasoning, yet it currently lacks a rigorous mathematical foundation. In this paper, we propose operads, mathematical structures that model many-in, one-out operations and compositions thereof, as a natural framework for describing question decomposition. We define the questions operad $Q$, in which operations correspond to question templates and composition corresponds to substitution of sub-answers, and show how QA models can be interpreted as algebras over $Q$. Beyond reframing existing practice, this operadic perspective points toward new methods, in particular a notion of operadic consistency, which measures whether a QA model's answers agree across the partial collapses of a question decomposition tree. Empirical evaluation of operadic consistency is reported in our companion paper (Bottman, Liu, and Richardson, 2026), which finds it strongly correlated with accuracy across twelve LLMs and four multi-hop QA datasets and outperforming standard temperature-based self-consistency baselines. We argue that operads are the natural mathematical home for question decomposition, and that invariants such as operadic consistency open new directions for analyzing and improving the reliability of multi-step reasoning.

14.
arXiv (CS.CL) 2026-06-17

ZeroSyl: Simple Zero-Resource Syllable Tokenization for Spoken Language Modeling

Pure speech language models aim to learn language directly from raw audio without textual resources. A key challenge is that discrete tokens from self-supervised speech encoders result in excessively long sequences, motivating recent work on syllable-like units. However, methods like Sylber and SyllableLM rely on intricate multi-stage training pipelines. We propose ZeroSyl, a simple training-free method to extract syllable boundaries and embeddings directly from a frozen WavLM model. Using L2 norms of features in WavLM's intermediate layers, ZeroSyl achieves competitive syllable segmentation performance. The resulting segments are mean-pooled, discretized using K-means, and used to train a language model. ZeroSyl outperforms prior syllabic tokenizers across lexical, syntactic, and narrative benchmarks. Scaling experiments show that while finer-grained units are beneficial for lexical tasks, our discovered syllabic units exhibit better scaling behavior for syntactic modeling.

15.
arXiv (CS.CV) 2026-06-17

Training LLMs with Reinforcement Learning over Digital Twin Representations for Reasoning-Intensive Surgical VideoQA

Surgical video question answering requires multi-step reasoning across semantic, spatial, and temporal dimensions. Existing methods architecturally compress videos into discrete token representations and couple visual perception with reasoning. This approach fragments continuous spatial-temporal relationships and has been shown to restrict multi-step reasoning capabilities. We introduce a reinforcement learning (RL) framework that trains large language models (LLMs) to decouple perception from reasoning by operating over digital twin representations constructed from surgical foundation models. Additionally, we introduce hierarchical representations across frame, temporal window, and procedure levels with probabilistic uncertainty estimates. Finally, we propose a novel reward that combines format validation with accuracy assessment through clinical plausibility evaluation and uncertainty-aware calibration for training. To demonstrate the capabilities of this approach, we introduce REAL-Colon-Reason, a colonoscopic benchmark with 2000 question-answer pairs across three complexity levels. We achieve state-of-the-art performance on REAL-Colon-Reason and two existing surgical VideoQA benchmarks REAL-Colon-VQA and EndoVis18-VQA.

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

iTryOn: Mastering Interactive Video Virtual Try-On with Spatial-Semantic Guidance

Video Virtual Try-On (VVT) aims to seamlessly replace a garment on a person in a video with a new one. While existing methods have made significant strides in maintaining temporal consistency, they are predominantly confined to non-interactive scenarios where models merely showcase garments. This limitation overlooks a crucial aspect of real-world apparel presentation: active human-garment interaction. To bridge this gap, we introduce and formalize a new challenging task: Interactive Video Virtual Try-On (Interactive VVT), where subjects in the video actively engage with their clothing. This task introduces unique challenges beyond simple texture preservation, including: (1) resolving the semantic ambiguity of interactions from standard pose information, and (2) learning complex garment deformations from video where interactive moments are sparse and brief. To address these challenges, we propose iTryOn, a novel framework built upon a large-scale video diffusion Transformer. iTryOn pioneers a multi-level interaction injection mechanism to guide the generation of complex dynamics. At the spatial level, we introduce a garment-agnostic 3D hand prior to provide fine-grained guidance for precise hand-garment contact, effectively resolving spatial ambiguity. At the semantic level, iTryOn leverages global captions for overall context and time-stamped action captions for localized interactions, synchronized via our novel Action-aware Rotational Position Embedding (A-RoPE). Extensive experiments demonstrate that iTryOn not only achieves state-of-the-art performance on traditional VVT benchmarks but also establishes a commanding lead in the new interactive setting, marking a significant step towards more dynamic and controllable virtual try-on experiences.

17.
arXiv (CS.AI) 2026-06-15

DiffusionBlocks: Block-wise Neural Network Training via Diffusion Interpretation

arXiv:2506.14202v4 Announce Type: replace-cross Abstract: End-to-end backpropagation requires storing activations throughout all layers, creating memory bottlenecks that limit model scalability. Existing block-wise training methods offer means to alleviate this problem, but they rely on ad-hoc local objectives and remain largely unexplored beyond classification tasks. We propose $DiffusionBlocks$, a principled framework for transforming transformer-based networks into genuinely independent trainable blocks that maintain competitive performance with end-to-end training. Our key insight leverages the fact that residual connections naturally correspond to updates in a dynamical system. With minimal modifications to this system, we can convert the updates to those of a denoising process, where each block can be learned independently by leveraging the score matching objective. This independence enables training with gradients for only one block at a time, thereby reducing memory requirements in proportion to the number of blocks. Our experiments on a range of transformer architectures (vision, diffusion, autoregressive, recurrent-depth, and masked diffusion) demonstrate that DiffusionBlocks training matches the performance of end-to-end training while enabling scalable block-wise training on practical tasks beyond small-scale classification. DiffusionBlocks provides a theoretically grounded approach that successfully scales to modern generative tasks across diverse architectures. Code is available at https://github.com/SakanaAI/DiffusionBlocks .

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

Bayesian Anytime Pareto Set Identification for Multi-Objective Multi-Armed Bandits

arXiv:2606.18785v1 Announce Type: cross Abstract: Identifying Pareto optimal solutions is critical to support multi-objective decision-making. We introduce the first anytime Multi-Objective Multi-Armed Bandit algorithm for the Pareto Set Identification problem, taking a Bayesian approach: Top-Two Pareto Front Thompson Sampling (TTPFTS). We benchmark TTPFTS against state-of-the-art fixed-budget Pareto Set Identification algorithms on synthetic environments. Next, we demonstrate its practical utility in a challenging multi-objective molecular discovery setting by efficiently exploring an ultra-large synthesis-on-demand molecular library. Furthermore, we introduce a novel uncertainty quantification metric that estimates our algorithm's confidence in the predicted Pareto set. We demonstrate that this metric effectively proxies true performance, yielding a robust methodology for monitoring learning progress in complex settings. Finally, we complement these empirical findings with a theoretical proof of the algorithm's asymptotic correctness.

19.
arXiv (CS.AI) 2026-06-17

Position: Modular Memory is the Key to Continual Learning Agents

arXiv:2603.01761v2 Announce Type: replace-cross Abstract: Foundation models have transformed machine learning through large-scale pretraining and increased test-time compute. Despite surpassing human performance in several domains, these models remain fundamentally limited in continuous operation, experience accumulation, and personalization, capabilities that are central to adaptive intelligence. While continual learning research has long targeted these goals, its historical focus on in-weight learning (IWL), i.e., updating a single model's parameters to absorb new knowledge, has rendered catastrophic forgetting a persistent challenge. Our position is that combining the strengths of In-Weight Learning (IWL) and the newly emerged capabilities of In-Context Learning (ICL) through the design of modular memory is the missing piece for continual adaptation at scale. We outline a conceptual framework for modular memory-centric architectures that leverage ICL for rapid adaptation and knowledge accumulation, and IWL for stable updates to model capabilities, charting a practical roadmap toward continually learning agents.

20.
arXiv (CS.CV) 2026-06-16

Latent Action Pretraining Through World Modeling

Vision-Language-Action (VLA) models have gained popularity for learning robotic manipulation tasks that follow language instructions. State-of-the-art VLAs, such as OpenVLA and $\pi_{0}$, were trained on large-scale, manually labeled action datasets collected through teleoperation. More recent approaches, including LAPA and villa-X, introduce latent action representations that enable unsupervised pretraining on unlabeled datasets by modeling abstract visual changes between frames. Although these methods have shown strong results, their large model sizes make deployment in real-world settings challenging. In this work, we propose LAWM, a model-agnostic framework to pretrain imitation learning models in a self-supervised way, by learning latent action representations from unlabeled video data through world modeling. These videos can be sourced from robot recordings or videos of humans performing actions with everyday objects. Our framework is able to transfer learned knowledge across tasks, environments, and embodiments. It outperforms models pretrained with ground-truth robot actions and other similar pretraining methods on the LIBERO benchmark and real-world setup, while being efficient and practical for real-world settings.

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

Learning to Decide with AI Assistance under Human-Alignment

arXiv:2605.12646v2 Announce Type: replace-cross Abstract: It is widely agreed that when AI models assist decision-makers in high-stakes domains by predicting an outcome of interest, they should communicate the confidence of their predictions. However, empirical evidence suggests that decision-makers often struggle to determine when to trust a prediction based solely on this communicated confidence. In this context, recent theoretical and empirical work suggests a positive correlation between the utility of AI-assisted decision-making and the degree of alignment between the AI confidence and the decision-makers' confidence in their own predictions. Crucially, these findings do not yet elucidate the extent to which this alignment influences the complexity of learning to make optimal decisions through repeated interactions. In this paper, we address this question in the canonical case of binary predictions and binary decisions. We first show that this problem is equivalent to a two-armed online contextual learning problem with full feedback, and establish a lower bound of $\Omega (\sqrt{|H| \cdot |B| \cdot T} )$ on the expected regret any learner can attain, where $H$ and $B$ denote the sets of human and AI confidence values. We then demonstrate that, under perfect alignment between AI and human confidence, a learner can attain an expected regret of $O(\sqrt{|H| \cdot T\log T})$ and, when $\sqrt{|H|} = O(\log T)$ and $B$ is countable, a non-trivial generalization of the Dvoretzky-Kiefer-Wolfowitz inequality improves the regret bound to $O(\sqrt{T\log T})$. Taken together, these results reveal that alignment can reduce the complexity of learning to make decisions with AI assistance. Experiments on real data from two different human-subject studies where participants solve simple decision-making tasks assisted by AI models show that our theoretical results are robust to violations of perfect alignment.

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

Retrieve, Don't Retrain: Extending Vision Language Action Models to New Tasks at Test Time

arXiv:2606.15631v1 Announce Type: cross Abstract: Extending a vision-language-action (VLA) policy to a new task typically requires task-specific teleoperated demonstrations and per-task fine-tuning, making adaptation costly in both data collection and compute. In this paper, we show that this target-side per-task adaptation cost can be replaced by retrieval. Our retrieval-augmented policy is trained once on paired demonstrations from the target embodiment (query) and a cheaper embodiment (pool, e.g., human-hand video), then frozen. New tasks are added at deployment by appending pool-side demonstrations to a retrieval pool. The frozen policy conditions on retrieved trajectories at every control step, so new tasks are absorbed by indexing data rather than updating parameters. Fine-tuning is needed only to take on a new, unseen embodiment, not for each new task. We show that retrieval improves policies beyond a specific backbone, including standard VLA policies, but its effect is especially pronounced in Cosmos Policy, a video-generation-based world-action model (WAM). In this setting, retrieval supplies coarse task progression, while the WAM's future-image objective provides an additional visual consistency signal that strengthens the retrieval-conditioned actions. On PushT, we study how retrieval provides a reusable high-level motion prior for cross-embodiment generalization to unseen goal angles, while on RoboTwin 2.0 our method outperforms cross-embodiment baselines on unseen tasks, and we additionally demonstrate the method on a real robot.

24.
arXiv (CS.AI) 2026-06-15

A Benchmark and Framework for Evaluating Next Action Predictions in Spreadsheets

arXiv:2606.13802v1 Announce Type: cross Abstract: Predictive code completion greatly accelerates how quickly developers work. In spreadsheets, despite being much more common, such auto-completion features are virtually non-existent. To address this gap, we introduce a benchmark for systems that observe a sequence of user actions in a spreadsheet and predict future actions. Two challenges are (1) the absence of edit histories in public spreadsheet corpora and (2) the complex space of spreadsheet actions (spatial, temporal, composite). To address (1), we manually curate 52 sequences of 12K actions that recreate spreadsheets from public corpora, seeded by parametrized heuristics and LLM refinement. To address (2), we propose an online evaluation that expects a prediction after each user action, accepts or rejects that prediction, updates the future actions upon acceptance, and repeats this until the target spreadsheet is obtained. We use multiple baseline predictors (including zero-shot LLMs, fine-tuned SLMs, and classical models) and analyze different properties that our benchmark teaches us, including but not limited to: properties of saved actions and false positives, efficiency, effect of user profiles, effect of triggers, and effect of context.

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

Rethinking the Role of Efficient Attention in Hybrid Architectures

Modern language models increasingly adopt hybrid architectures that combine full attention with efficient attention modules, such as sliding-window attention (SWA) and recurrent sequence mixers. However, how these efficient modules shape model capabilities remains poorly understood. To address this gap, we conduct a systematic analysis across hybrid architectures from three perspectives: scaling behavior, mechanism analysis, and architecture design. First, from a scaling perspective, we find that efficient-attention design primarily affects how fast long-context capability emerges, while different hybrids eventually converge to comparable long-context performance under sufficient training. Second, mechanistically, we show that long-range retrieval is mainly carried by full attention, whereas efficient attention shapes its optimization trajectory. This explains a counter-intuitive phenomenon we call Large-Window Laziness: larger SWA windows can delay the formation of retrieval heads in full-attention layers. Third, guided by this mechanism, we show that applying NoPE to only the full-attention layers of a small-window SWA hybrid substantially improves long-context performance with negligible impact on short-context performance.