# Research Papers — 2026-05 · bundle page 65 of 326 > Full-text markdown of papers #3201–3250 of 16266 published in 2026-05, newest first. AI-readable mirrors of the full 16266-paper 2026-05 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-05/page/65 [← Previous bundle](https://aipod.dev/papers/archive/2026-05/page/64.md) [Next bundle →](https://aipod.dev/papers/archive/2026-05/page/66.md) ### FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies Authors: Xintong Hu, Xuhong Huang, Jinyu Zhang, Yutong Yao, Yuchong Sun, Qiuyue Wang, Mingsheng Li, Sicheng Xie, Yitao Liu, Junhao Chen, Yixuan Chen, Yingming Zheng et al. Published: 2026-05-26 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2605.27284v3 Vision-Language-Action (VLA) models are increasingly expected to not only complete robot tasks, but also follow human instructions about how those tasks should be executed. However, existing robot datasets usually pair trajectories with coarse goal-level language, leaving execution-critical details such as active arm, approach direction, and contact region unspecified. This limits steerable policy learning and robotic video understanding. We introduce FineVLA, an open framework for action-aligned fine-grained VLA supervision. The framework includes: (1) a data construction tool that unifies 972,247 trajectories across 85K tasks from 10 open-source robot datasets and builds FineVLA-Data, a human-verified dataset of 47,159 fine-grained trajectories; (2) a held-out benchmark with 500 videos, 11,631 atomic facts, and 1,030 VQA questions; (3) a robotics-specialized VLM annotator for scalable fine-grained annotation; and (4) a steerable VLA policy trained with controlled mixtures of fine-grained and raw goal-level instructions. Our experiments yield three findings. First, fine-grained supervision does not sacrifice goal-level success: FG-only improves over Raw-only by +1.4 to +8.1 success-rate points across settings. Second, fine-grained and raw instructions are complementary, following a consistent inverted-U trend peaking at FG:Raw = 1:2 to 1:1. The best mixed setting reaches 86.8%/82.5% in RoboTwin simulation and 62.7/100 in real-world dual-arm manipulation (vs. 49.9 Raw-only). Third, fine-grained supervision improves steerable control: the largest real-world gains appear on pose (+23), color (+18), and approach direction (+18)--factors where goal-level instructions provide no guidance. Overall, fine-grained language should augment goal-level instructions: specifying how to execute alongside what to achieve. Project page: https://finevla.xlang.ai/ --- ### FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies Authors: Xintong Hu, Xuhong Huang, Jinyu Zhang, Yutong Yao, Yuchong Sun, Qiuyue Wang, Mingsheng Li, Sicheng Xie, Yitao Liu, Junhao Chen, Yixuan Chen, Yingming Zheng et al. Published: 2026-05-26 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2605.27284v2 Vision-Language-Action (VLA) models are increasingly expected to not only complete robot tasks, but also follow human instructions about how those tasks should be executed. However, existing robot datasets usually pair trajectories with coarse goal-level language, leaving execution-critical details such as active arm, approach direction, and contact region unspecified. This limits steerable policy learning and robotic video understanding. We introduce FineVLA, an open framework for action-aligned fine-grained VLA supervision. The framework includes: (1) a data construction tool that unifies 972,247 trajectories across 85K tasks from 10 open-source robot datasets and builds FineVLA-Data, a human-verified dataset of 47,159 fine-grained trajectories; (2) a held-out benchmark with 500 videos, 11,631 atomic facts, and 1,030 VQA questions; (3) a robotics-specialized VLM annotator for scalable fine-grained annotation; and (4) a steerable VLA policy trained with controlled mixtures of fine-grained and raw goal-level instructions. Our experiments yield three findings. First, fine-grained supervision does not sacrifice goal-level success: FG-only improves over Raw-only by +1.4 to +8.1 success-rate points across settings. Second, fine-grained and raw instructions are complementary, following a consistent inverted-U trend peaking at FG:Raw = 1:2 to 1:1. The best mixed setting reaches 86.8%/82.5% in RoboTwin simulation and 62.7/100 in real-world dual-arm manipulation (vs. 49.9 Raw-only). Third, fine-grained supervision improves steerable control: the largest real-world gains appear on pose (+23), color (+18), and approach direction (+18)--factors where goal-level instructions provide no guidance. Overall, fine-grained language should augment goal-level instructions: specifying how to execute alongside what to achieve. Project page: https://finevla.xlang.ai/ --- ### Causal Risk Minimization for High-Dimensional Treatments Authors: Nikita Dhawan, Arnav Paruthi, Andrew Kim, Lovedeep Gondara, Jekaterina Novikova, Chris J. Maddison Published: 2026-05-26 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2605.27281v1 Predicting the effect of interventions with many possible variations, e.g., therapeutic content that affects mental health outcomes or an earnings call transcript that drives movement in share price, is useful across several domains. However, classical causal estimators tend to assume that all possible interventions are observed, which is infeasible when interventions vary widely, for instance, in the space of all text strings. We adapt a well-known approach of recasting causal inference as a learning problem, to address high-dimensional treatment spaces. Specifically, under standard assumptions like no unobserved confounding, we show that causal error decomposes into a series of moment-balancing errors of increasing order, and design objectives that directly improve causal estimation. We also show how to project the effect of a high-dimensional treatment onto lower-dimensional treatment attributes, which allows a single model to answer several causal questions without additional attribute-specific training. We empirically evaluate our estimators in settings with high-dimensional continuous, discrete, and text treatments, the last of which used a semi-synthetic dataset of Amazon Reviews. Our experiments demonstrate the benefit of higher-order balance error optimization and competitive performance of projected causal estimates with attribute-specific estimators. --- ### SIA: Self Improving AI with Harness & Weight Updates Authors: Prannay Hebbar, Yogendra Manawat, Samuel Verboomen, Alesia Ivanova, Selvam Palanimalai, Kunal Bhatia, Vignesh Baskaran Published: 2026-05-26 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2605.27276v2 Humans are the bottleneck in building and improving AI. Both the models and the agents that wrap them are written, tuned, and corrected by people. The long-horizon goal of an AI that can figure out how to improve itself remains open. Two largely disjoint research lines attack this bottleneck. The harness-update school has a meta-agent rewrite the scaffold of a task-specific agent (its tools, prompts, retry logic, and search procedure) while the model weights are held fixed. The test-time training school uses hand-written RL pipelines to update the model's own weights on task feedback while the harness is held fixed. These two silos operate in isolation. We propose SIA, a self-improving loop in which a language-model agent (the Feedback-Agent) updates both the harness and the weights of a task-specific agent. We evaluate across three contrasting domains: Chinese legal charge classification, low-level GPU kernel optimisation, and single-cell RNA denoising. Combining both levers outperforms scaffold iteration alone on all three benchmarks. SIA-W+H achieves 25.1% over prior SOTA on LawBench, 12.4% faster GPU kernels than prior SOTA (1,017 vs 1,161 μs), and 20.4% over prior SOTA on denoising. Harness updates make the model agentic, shaping how it searches and acts, while weight updates build the domain intuition that no prompt or scaffold can instil. --- ### Inverse Control Constrained Optimization of Vessel Speed Decisions Under Environmental Risk: Evidence from Arctic Shipping Authors: Mauli Pant, Linda Fernandez, Indranil Sahoo Published: 2026-05-26 Categories: stat.AP, stat.ML arXiv: https://arxiv.org/abs/2605.27270v1 Understanding how decision makers balance operational efficiency with environmental and ecological risks is central to vessel navigation. We model vessel speed as a control variable in a constrained optimization framework in which vessel operators balance multiple competing objectives, including transit efficiency, ice related navigational risk, and whale related ecological risk. The underlying risk parameters are estimated using over 14 million Automatic Identification System (AIS) observations from the United States Arctic (2010-2019), together with environmental covariates and spatially explicit whale density estimates. The framework incorporates a nonlinear risk objective, vessel heterogeneity, and regularization to ensure stable and interpretable results. The inferred trade offs reveal distinct decision making patterns across vessel groups and navigational statuses. Vessel types such as Tug Tow and Cargo balance operational speed with environmental and ecological considerations. In contrast, several vessel groups, including Fishing, Passenger, and Unspecified vessels, are strongly influenced by ice related risk, while Pleasure Craft and Tankers exhibit higher sensitivity to whale related risk. Across navigational status categories, similar heterogeneity is observed. The dominant status, under way using engine, displays a clear trade off, whereas other statuses, such as aground and undefined, are strongly shaped by ice related constraints. Statuses including restricted maneuverability and engaged in fishing exhibit higher estimated sensitivity to whale related risk, though with substantial uncertainty. Sensitivity analysis indicates that increasing whale-related risk weighting produces limited changes in model-implied optimal speed, whereas increasing ice-related risk leads to more consistent reductions. --- ### Transfer Learning using 66 Diseases for Disease Forecasting Applications Authors: Lauren J Beesley, Alexander C Murph, Dave Osthus, Lauren A Castro Published: 2026-05-26 Categories: cs.LG, stat.AP arXiv: https://arxiv.org/abs/2605.27269v1 Disease forecasting models typically rely on a single data stream, making models brittle when histories are short or noisy. Recent top-performing models have shown that synthesizing multiple reporting systems for the same disease improves performance. Other recent work takes this idea a step further, using transfer learning to train a forecasting model for one disease using data from a different disease. We expand upon each of these approaches greatly, training machine learning models on data that span 66 infectious diseases and several data streams. We investigate the value of incorporating different data streams for forecasting 20 different disease data streams. We find that incorporating other data streams improves forecasting in the vast majority (84.9%) of time series and model structures considered. However, our work highlights that the quality of the added data matters, where adding data extremely different from the target data stream can sometimes degrade forecast performance. A major contribution of this work is in compiling a publicly-available database of data for use by the infectious disease forecasting community. --- ### Lost in Sampling: Assessing Lexical Reachability in LLMs via the Word Coverage Score (WCS) Authors: Samer Awad, Javier Conde, Carlos Arriaga, Tairan Fu, Javier Coronado-Blázquez, Pedro Reviriego Published: 2026-05-26 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.27268v2 Modern Large Language Models (LLMs) are often criticized for producing repetitive and homogeneous text, despite possessing vast latent vocabularies. While previous research has focused on model knowledge and training data, we investigate the role of decoding mechanics in suppressing linguistic diversity. We introduce the Word Coverage Score (WCS), a metric that quantifies the extent to which contextually appropriate human vocabulary is mathematically pruned by standard sampling filters (e.g., Top-$p$, Top-$k$, and Min-$p$). Rather than assessing static knowledge, the WCS measures the lexical survival rate of low-frequency, high-information human words as a function of sampling parameters. By auditing open-weight models on human-authored corpus fragments, we identify which logical lexical choices are rendered unreachable by the decoder, even when they reside within the probability space. Our results provide quantitative evidence that industry-standard sampling defaults act as unintended censorship mechanisms, smoothing the unique textures of human expression into a homogenized discourse. The WCS offers a rigorous framework for optimizing the trade-off between text coherence and lexical richness, providing a diagnostic tool for preserving the diversity of human language in generative models. --- ### Lost in Sampling: Assessing Lexical Reachability in LLMs via the Word Coverage Score (WCS) Authors: Samer Awad, Javier Conde, Carlos Arriaga, Tairan Fu, Javier Coronado-Blázquez, Pedro Reviriego Published: 2026-05-26 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.27268v1 Modern Large Language Models (LLMs) are often criticized for producing repetitive and homogeneous text, despite possessing vast latent vocabularies. While previous research has focused on model knowledge and training data, we investigate the role of decoding mechanics in suppressing linguistic diversity. We introduce the Word Coverage Score (WCS), a metric that quantifies the extent to which contextually appropriate human vocabulary is mathematically pruned by standard sampling filters (e.g., Top-$p$, Top-$k$, and Min-$p$). Rather than assessing static knowledge, the WCS measures the lexical survival rate of low-frequency, high-information human words as a function of sampling parameters. By auditing open-weight models on human-authored corpus fragments, we identify which logical lexical choices are rendered unreachable by the decoder, even when they reside within the probability space. Our results provide quantitative evidence that industry-standard sampling defaults act as unintended censorship mechanisms, smoothing the unique textures of human expression into a homogenized discourse. The WCS offers a rigorous framework for optimizing the trade-off between text coherence and lexical richness, providing a diagnostic tool for preserving the diversity of human language in generative models. --- ### Kan Extension Transformers: A Categorical Unification of Attention, Diffusion, and Predict-Detach Self-Conditioning Authors: Sridhar Mahadevan Published: 2026-05-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.27259v2 We propose Kan Extension Transformers (KETs) as a categorical design language for a diverse group of Transformer implementations. A layer can be viewed generally as a weighted structured extension operator: attention uses token neighborhoods, geometric mixing uses sparse incidences, and KET uses simplicial sources. This operator is an actual enriched left Kan extension only when the source values are functorial, the weights are representable hom-objects (or the specified profunctor action), and aggregation realizes the corresponding coend; otherwise ``Kan-style'' denotes an interpretation rather than an identity theorem. Predict-detach blocks gradients through a predictive carrier and avoids transporting teacher-forced hidden states, but detach alone does not make a noncausal update strictly autoregressive: every carrier consumed at target $t$ must also be measurable from the prefix available at $t$. We evaluate 12 implementations on Penn Treebank, WikiText-2, and WikiText-103 across strict-causal and self-conditioned regimes, using widths $d=64,256$ and depths $L=2,8,16$ across the reported studies. Quadratic KET is strongest among the compared strict-causal architectures on WikiText-2 and WikiText-103; the largest cross-regime gains arise from additional self-conditioning information, not neighborhood design alone. --- ### PilotTTS: A Disciplined Modular Recipe for Competitive Speech Synthesis Authors: Bowen Li, Shaotong Guo, Zhen Wang, Yang Xiang, Mingli Jin, Yihang Lin, Jiahui Zhao, Weibo Xiong, Dongrui Zhang, Keming Chen, Yunze Gao, Zeyang Lin et al. Published: 2026-05-26 Categories: cs.SD, cs.AI arXiv: https://arxiv.org/abs/2605.27258v2 Building state-of-the-art text-to-speech (TTS) systems typically demands millions of hours of proprietary data and complex multi-stage architectures, creating substantial barriers for resource-constrained research teams. In this report, we present PilotTTS, a lightweight autoregressive TTS system that achieves competitive performance through minimalist architecture and rigorous data engineering. PilotTTS is trained on only 200K hours of data processed entirely with open-source tools. Specifically, our contributions are: (1) a reproducible multi-stage data processing pipeline covering quality assessment, label annotation, and filtering, and (2) a compact model architecture that employs Q-Former-based conditioning to decouple speaker identity from speaking style via cross-sample paired training. Within a unified framework, PilotTTS supports zero-shot voice cloning, emotion synthesis (11 categories), paralinguistic synthesis (4 categories), and Chinese dialect synthesis (14 dialects). On the Seed-TTS Eval benchmark, PilotTTS achieves the lowest WER of 1.50% on test-en, a CER of 0.87% on test-zh, and the highest speaker similarity on both test sets (0.862 and 0.815), outperforming systems trained on significantly larger datasets. We release the complete data pipeline recipe, pretrained weights, and code at https://github.com/AMAPVOICE/PilotTTS. --- ### Pair-In, Pair-Out: Latent Multi-Token Prediction for Efficient LLMs Authors: Wenhui Tan, Minghao Li, Xiaoqian Ma, Siqi Fan, Xiusheng Huang, Liujie Zhang, Ruihua Song, Weihang Chen Published: 2026-05-26 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.27255v2 Long chain-of-thought reasoning has made autoregressive decoding the dominant inference cost of modern large language models. Existing methods target either the input side (latent compression) or the output side (speculative decoding and multi-token prediction, MTP), but the two lines of work have been pursued independently. Moreover, output-side methods must incur an expensive verifier pass to validate the unreliable draft tokens predicted by MTP. To address these issues, we propose \textbf{Pair-In, Pair-Out (PIPO)}, which unifies both sides by viewing a latent compressor and an MTP head as mirror-image operations: the compressor folds two input tokens into one latent representation, while the MTP head unfolds one hidden state into one additional output token. To remove the verifier cost without sacrificing reliability, PIPO trains a lightweight confidence head that decides whether draft tokens should be accepted. We observe that On-Policy Distillation (OPD) naturally matches the rejection-sampling criterion of speculative decoding, so the confidence head can be trained alongside OPD with negligible extra cost. Experiments on AIME 2025, GPQA-Diamond, LiveCodeBench v6, and LongBench v2 with Qwen3.5-4B and 9B backbones show that PIPO improves pass@4 over regular decoding by up to $+7.15$ points, while delivering up to $2.64\times$ first-token-latency and $2.07\times$ per-token-latency speedups. Project Page: GitHub.com/RedAI-Infra/PIPO. --- ### LUCoS: Latent Unsupervised Context Selection for Tabular Foundation Models Authors: Oroel Ipas, Guillermo Gomez-Trenado, Rocío Romero-Zaliz, Isaac Triguero Published: 2026-05-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.27254v1 Selecting which instances to label is a key challenge in low-label tabular learning. For recent Tabular Foundation Models such as TabPFN, context selection directly determines predictive performance. Supervised oracle experiments show that carefully chosen labeled context sets can strongly outperform random selection under the same labeling budget. However, the cold-start setting, where instances must be selected before any labels are available, has received little attention in the TFM literature. This problem is fundamentally geometric. In vision and language, foundation models induce embedding spaces where simple geometric selection methods are effective. In contrast, tabular instance selection has so far been performed predominantly in the original tabular space, which lacks a natural metric; heterogeneous types, mixed scales, and nonlinear interactions make raw-space distances unreliable for context construction, and original-space selection falls below random on the majority of datasets as the budget grows. We propose LUCoS (Latent Unsupervised Context Selection), which replaces raw-feature geometry with the latent geometry induced by embeddings from an unsupervised Prior-Fitted Network (PFN) and selects representative medoids as context. Evaluated on 67 OpenML-CC18 datasets across six low-label budgets, LUCoS ranks first under mean AUC, ACC, and F1, with conclusions stable across metrics and dataset-level robustness checks. A gain decomposition reveals a simple mechanism: at the smallest budgets, the main benefit comes from enforcing coverage; as the budget increases, the decisive factor becomes the representation space in which coverage is measured. LUCoS mitigates failures of original feature space selection, showing that reliable unsupervised context selection depends less on selector sophistication than on defining representativeness in a meaningful representation geometry. --- ### Gumbel Machine: Counterfactual Student Writing Generation via Gumbel Noise Steering Authors: Hunter McNichols, Alexander Scarlatos, Mihai Dascalu, Danielle McNamara, Andrew Lan Published: 2026-05-26 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2605.27249v1 An effective method of teaching across disciplines is to provide examples of high-quality work. However, an example may be significantly different from a student's current work, making it challenging for them to emulate. An ideal learning demonstration is a counterfactual version of the student work, an improved version that is still similar to their own. Existing automated approaches for counterfactual text generation using Large Language Models (LLMs) result in domain-specific systems that are difficult to translate into practical applications. We present the Gumbel Machine, a flexible, modular approach to generating counterfactuals that leverages LLM instruction-following capabilities while encouraging similarity to a reference factual text. Central to our approach is a novel, controlled decoding algorithm, $β$-Hindsight control, which uses latent randomness as a tunable similarity control mechanism during counterfactual generation. Experiments on datasets of student writing, scored on various criteria, demonstrate the effectiveness of our approach at generating counterfactuals both rubric-consistent and similar to a reference. --- ### Many Logics, One Methodology: A Plea for Logical Pluralism in Formalised Reasoning (preprint) Authors: Christoph Benzmüller, Daniel Kirchner, Luca Pasetto Published: 2026-05-26 Categories: cs.LO, cs.AI, math.LO arXiv: https://arxiv.org/abs/2605.27246v1 This position statement looks back on two decades of work on shallow embeddings of non-classical logics in classical higher-order logic (HOL), a line of research that expanded into a range of logic embeddings in HOL and inspired the LogiKEy logic-pluralistic knowledge representation and reasoning methodology. This paper advances the case for logical pluralism at object-logic level within a unifying meta-logical framework such as LogiKEy, grounding the argument in computational metaphysics. More broadly, it advocates principled support for logical pluralism in modern proof assistants, and cautions against logical imperialism -- the rigid adoption of a single foundational logic for large-scale theory developments -- which impedes the interdisciplinary reuse that LogiKEy is designed to enable. --- ### Symbolic Regression via Latent Iterative Refinement Authors: Xieting Chu, Sriram Vishwanath, Vijay Ganesh Published: 2026-05-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.27245v1 Symbolic regression (SR) seeks closed-form mathematical expressions that fit observed data. Neural SR methods amortize the search by training an encoder to map observations directly to expressions in a single pass, but this amortized inference leaves a residual amortization gap between its one-shot prediction and the true posterior. We propose Latent Equation Embedding (LEE), a framework that closes this gap through iterative amortized inference in a functionally grounded latent space. LEE learns a shared latent space Z equipped with three components: an encoder f_theta that jointly embeds symbolic tokens and numerical observations into a single latent vector z; an expression decoder g_expr that reconstructs formulas from z; and an evaluation decoder g_eval that predicts function values from z, explicitly grounding the latent space in functional behavior. At inference, LEE performs iterative refinement by re-encoding decoded expressions jointly with observations, progressively improving the latent estimate. LEE uses the encoder itself as a learned inference optimizer: each re-encoding step implicitly computes the mismatch between the candidate and the data. Because g_eval is differentiable in z, we additionally interleave continuous gradient descent with discrete re-encoding, yielding a hybrid iterative and gradient refinement procedure. On SRBench across three noise levels, against 19 baselines spanning genetic programming, symbolic-neural hybrids, and pre-trained Transformers, LEE produces expressions 2--10x simpler than the strongest accuracy-oriented baselines, including Operon, GP-GOMEA, TPSR, RAG-SR, and GenSR, with complexity 8--11 versus 20--90. These results advance the low-complexity region of the accuracy-complexity Pareto frontier and show graceful degradation as noise increases. --- ### Can Retrieval Heads See Images? Multimodal Retrieval Heads in Long-Context Vision-Language Models Authors: Aaron Branson Cigres Li, Zhaowei Wang, Yu Zhao, Yiming Du, Haobo Li, Xiyu Ren, Ginny Wong, Simon See, Lishu Luo, Haodong Duan, Pasquale Minervini, Yangqiu Song Published: 2026-05-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.27243v3 Large vision-language models increasingly rely on long-context modeling to reason over documents, hour-level videos, and long-horizon agent trajectories, requiring them to locate relevant evidence across interleaved text and images. Prior work has studied this behavior using retrieval heads in large language models, but its copy-based criterion does not directly apply when evidence appears in images. We introduce a multimodal retrieval head detection method that scores attention from question tokens to textual or visual evidence. With this method, we show that multimodal retrieval heads are sparse, intrinsic, and causally important: only 4.4-10.2% of attention heads account for 50% of the positive retrieval-score mass, and masking the top-5% selected heads reduces performance on two tasks from MMLongBench by 42.5 and 62.3 percentage points, while random-head masking is far less damaging. Further analysis shows that these heads are partly shared across modalities yet remain dynamic within each modality, with image retrieval heads changing more than text retrieval heads as context length and haystack modality change. Without further training, we find that these heads can also be used directly to rank visually rich documents: on MMDocIR, Qwen3-VL-8B selected-head scoring improves Recall@1 by 7.7/7.4 macro/micro points for page retrieval and 6.3/6.8 points for layout retrieval over the strongest reported baseline. --- ### Can Retrieval Heads See Images? Multimodal Retrieval Heads in Long-Context Vision-Language Models Authors: Aaron Branson Cigres Li, Zhaowei Wang, Yu Zhao, Yiming Du, Haobo Li, Xiyu Ren, Ginny Wong, Simon See, Lishu Luo, Haodong Duan, Pasquale Minervini, Yangqiu Song Published: 2026-05-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.27243v2 Large vision-language models increasingly rely on long-context modeling to reason over documents, hour-level videos, and long-horizon agent trajectories, requiring them to locate relevant evidence across interleaved text and images. Prior work has studied this behavior using retrieval heads in large language models, but its copy-based criterion does not directly apply when evidence appears in images. We introduce a multimodal retrieval head detection method that scores attention from question tokens to textual or visual evidence. With this method, we show that multimodal retrieval heads are sparse, intrinsic, and causally important: only 4.4-10.2% of attention heads account for 50% of the positive retrieval-score mass, and masking the top-5% selected heads drops MMLongBench-Doc from 48.2% to 5.7% and SlideVQA from 71.2% to 8.9%, while random-head masking is far less damaging. Further analysis shows that these heads are partly shared across modalities yet remain dynamic within each modality, with image retrieval heads changing more than text retrieval heads as context length and haystack modality change. Without further training, we find that these heads can also be used directly to rank visually rich documents: on MMDocIR, Qwen3-VL-8B selected-head scoring improves Recall@1 by 7.7/7.4 macro/micro points for page retrieval and 6.3/6.8 points for layout retrieval over the strongest reported baseline. --- ### ENPMR-Bench: Benchmarking Proactive Memory Retrieval for Emotional Support Agents Authors: Xing Fu, Yulin Hu, Mengtong Ji, Haozhen Li, Yixin Sun, Weixiang Zhao, Yanyan Zhao, Bing Qin Published: 2026-05-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.27240v1 Memory-augmented language agents are increasingly deployed in affective applications such as emotional support, where understanding and responding to users' latent emotional needs is critical. However, existing research often treats memory as a tool for factual retrieval, overlooking its role in shaping users' emotional experiences. In this work, we introduce ENPMR-Bench, a benchmark for evaluating Emotional Need-aware Proactive Memory Retrieval (ENPMR), a core capability that enables agents to infer users' latent emotional needs and proactively retrieve appropriate memories to support empathetic interaction. Grounded in Maslow's hierarchy of needs, ENPMR-Bench includes over 1,800 memory-augmented dialogues and defines structured mappings between emotional needs and supportive memory types. Experimental results demonstrate that current retrieval paradigms, including both embedding-based and LLM-driven approaches, exhibit substantial deficiencies, with empathy scores significantly lagging behind golden memory conditions. While chain-of-thought prompting improves the alignment between inferred emotional needs and retrieved memories to some extent, a notable performance gap remains. Together, these findings reveal critical limitations in current agents and outline directions for advancing personalized emotional support through need-sensitive memory retrieval. --- ### Temporal Simultaneity Predicts Annotation Quality in Sentiment Corpora Authors: Idris Abdulmumin, Mokgadi Penelope Matloga, Tadesse Destaw Belay, Botshelo Kondowe, Letlhogonolo Mohleleng, Hareaipha Nkopo Letsoalo, Shamsuddeen Hassan Muhammad, Vukosi Marivate Published: 2026-05-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.27239v1 Annotation quality is difficult to sustain when campaigns span weeks or months with small annotator pools. We present a Setswana sentiment dataset of 3,565 tweets annotated by three native-speaker annotators across eight batches and examine why inter-annotator agreement (IAA) declines over time. Despite an aggregate Randolph's free-marginal Kappa of $κ= 0.76$, "excellent," per-batch $κ$ falls by more than 32 points across the annotation task. Through six targeted analyses, we find that (i) label confusion concentrates on the negative/neutral boundary, (ii) two annotators show run-length drift consistent with autopilot labeling, and (iii) the dominant predictor of $κ$ is temporal simultaneity: tweets labeled within one minute achieve $κ= 0.98$, while those labeled more than a day apart reach only $κ= 0.65$. Annotation speed and tweet-level linguistic features show no meaningful association with $κ$. We benchmark three open multilingual encoders and proprietary models (GPT-5 and Gemini) on three-class sentiment classification; fine-tuning yields gains of 29 to 43 macro-F1 points over pretrained baselines, with GPT-5 few-shot leading overall (62.2 macro-F1). We release the dataset, per-annotation timestamps, and analysis code to support reproducible quality auditing for future African language NLP resources. --- ### Explainable Comparison of Feature-Based and Deep Learning Models for TROPOMI Methane Plume Screening Authors: Solomiia Kurchaba, Joannes D. Maasakkers, Berend J. Schuit, Ilse Aben Published: 2026-05-26 Categories: cs.LG, physics.ao-ph arXiv: https://arxiv.org/abs/2605.27236v1 Continuous and global detection of large methane emissions is a crucial step for global warming mitigation. Satellite observations, such as from S5P/TROPOMI, combined with plume detection algorithms, can play a key role in this effort. However, not all TROPOMI plume detections that look like methane emission plumes are the result of actual emissions. A significant part of the plume-like features in the data are retrieval artifacts. Such artifacts could be the result of variations in elevation or albedo gradients, high concentrations of aerosols, coastal lines, water bodies, etc. Previous work approached the problem of plume-artifact classification by means of a Support Vector Machine Classifier (SVC), trained on an extensive set of observation-based scalar features designed by domain experts. However, such an approach limits the information scope received by the algorithm to what is deemed to be important by the experts, breaks the spatial relationship between pixels, and loses information during the process of statistical aggregation. In this study, we compare feature-based (SVC, Random Forest, XGBoost) and image-based (ResNet-18, ResNet-34) models for methane plume-artifact classification under balanced and imbalanced evaluation settings. To interpret the results, we apply SHAP-based explainability to both model families. Our findings provide practical guidance for model selection in operational methane-screening workflows such as the CAMS Methane Hotspot Explorer. --- ### MRT: Masked Region Transformer for Layered Image Generation and Editing at Scale Authors: Zhicong Tang, Zhao Zhang, Jingye Chen, Mohan Zhou, Yifan Pu, Yuchi Liu, Yalong Bai, Ethan Smith, Yuhui Yuan Published: 2026-05-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.27235v1 Layered image generation and editing is a fundamental capability that enables layer-wise reuse, editing, and composition of generated visual content, analogous to word-level editing in natural language. Despite its importance, this remains an underexplored area at scale. To address this gap, we present MRT, a 20B-parameter masked region diffusion model tailored for multi-layer transparent image generation and editing, trained on over 10M multilingual design samples spanning diverse aspect ratios and textual prompts. To fully leverage this scale, we make two key technical contributions. First, we unify three complementary tasks including text-to-layers, image-to-layers, and layers-to-layers within a shared masked region diffusion framework, where selective token masking enables flexible layer-wise generation and editing. Second, to enable overflow layer generation, we introduce an overflow-aware canvas layer that handles boundary inconsistencies and supports semi-transparent background synthesis, enabling complete editable layers extending beyond visible canvas boundaries. Additionally, we apply diffusion distillation to achieve 8-step, real-time multi-layer generation with minimal quality degradation. Extensive experiments demonstrate that our framework substantially outperforms prior state-of-the-art approaches, including various commercial systems, across all three tasks, establishing a new benchmark for multi-layer transparent image generation. Notably, our model significantly outperforms the concurrent Qwen-Image-Layered model in image-to-layers quality according to user-study results, while achieving 10-100\times faster inference and reducing activation GPU memory consumption by 50-90\% during image-to-layer inference. --- ### The Coverage Illusion: From Pre-retrieval Routing Failure to Post-retrieval Cascades in a Production RAG System Authors: Zafar Hussain, Kristoffer Nielbo Published: 2026-05-26 Categories: cs.CL, cs.IR arXiv: https://arxiv.org/abs/2605.27220v1 In modern RAG pipelines, query augmentation methods such as HyDE and query expansion are applied to every query, resulting in substantial LLM inference costs and increased end-to-end latency. The empirical justification for this overhead in real production traffic remains largely unexplored. We present a case study of the Danish National Encyclopedia, evaluating five retrieval workflows over 20,000 query-workflow pairs from production traffic and synthetic conditions. In this system, synthetic queries suggest that LLM augmentation is needed for over 90% of queries to achieve high retrieval coverage. However, under our production deferral policy, only 27.8% of real user queries need LLM augmentation. We call this gap the Coverage Illusion and attribute it to a structural mismatch between synthetic and real query distributions. Pre-retrieval routing cannot resolve this gap, as the need for LLM augmentation is only revealed after searching the index, a result confirmed by our evaluation of four machine learning paradigms. The coverage gap, undetectable from the query alone, motivates a post-retrieval cascade that runs workflows in cheapest-first order and escalates to LLM augmentation only when a step returns no documents. Operating entirely without training overhead or secondary serving infrastructure, the cascade improves quality by +0.140 Composite Overall points over Always-HyDE, reduces latency by 31.8%, and serves 72.2% of real user queries without LLM augmentation. --- ### Nonlinear Data Integration via Kernel Methods for Data Collaboration Analysis Authors: Yamato Suetake, Yuta Kawakami, Shunnosuke Ikeda, Yuichi Takano Published: 2026-05-26 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2605.27219v2 Collaborative analysis of decentralized confidential datasets is important, but direct sharing of original datasets is often restricted by privacy and institutional constraints. Data collaboration (DC) analysis transforms each dataset into privacy-preserving intermediate representations via party-specific obfuscation functions and integrates them into common collaboration representations using an anchor dataset. However, many existing DC analysis methods rely on linear transformations for data obfuscation and integration, which may increase reconstruction risk. Although nonlinear dimensionality reduction can mitigate this risk, conventional linear integration methods cannot accurately align intermediate representations produced by nonlinear transformations. Moreover, existing integration methods mainly minimize discrepancies among parties and do not explicitly incorporate geometric or target-variable information useful for downstream analysis. To overcome these limitations, we first formulate linear target-normalized integration (LTI) as a linear integration method and then kernelize it to obtain kernel-based target-normalized integration (KTI). KTI admits a globally optimal solution via kernel ridge regression and an eigenvalue problem. We also introduce graph regularization and a centering constraint so that the target representation can capture geometric and target-variable information useful for downstream analysis. Experiments on image classification tasks demonstrate that KTI improves classification accuracy over existing linear integration methods under nonlinear dimensionality reduction, with further gains from target-variable-aware graph regularization and centering. The results also show that dimensionality reduction choices substantially affect both classification accuracy and reconstruction risk. --- ### Nonlinear Data Integration via Kernel Methods for Data Collaboration Analysis Authors: Yamato Suetake, Yuta Kawakami, Shunnosuke Ikeda, Yuichi Takano Published: 2026-05-26 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2605.27219v1 Collaborative analysis of decentralized confidential datasets is important, but direct sharing of original datasets is often restricted by privacy and institutional constraints. Data collaboration (DC) analysis transforms each dataset into privacy-preserving intermediate representations via party-specific obfuscation functions and integrates them into common collaboration representations using an anchor dataset. However, many existing DC analysis methods rely on linear transformations for data obfuscation and integration, which may increase reconstruction risk. Although nonlinear dimensionality reduction can mitigate this risk, conventional linear integration methods cannot accurately align intermediate representations produced by nonlinear transformations. Moreover, existing integration methods mainly minimize discrepancies among parties and do not explicitly incorporate geometric or target-variable information useful for downstream analysis. To overcome these limitations, we first formulate linear kernel integration (LKI) as a linear integration method and then kernelize it to obtain nonlinear kernel integration (NKI). NKI admits a globally optimal solution via kernel ridge regression and an eigenvalue problem. We also introduce graph regularization and a centering constraint so that the target representation can capture geometric and target-variable information useful for downstream analysis. Experiments on image classification tasks demonstrate that NKI improves classification accuracy over existing linear integration methods under nonlinear dimensionality reduction, with further gains from target-variable-aware graph regularization and centering. The results also show that dimensionality reduction choices substantially affect both classification accuracy and reconstruction risk. --- ### Qiskit QuantumKatas: Adapting Microsoft's Quantum Computing exercises for LLM evaluation Authors: Juan Cruz-Benito, Ismael Faro Published: 2026-05-26 Categories: quant-ph, cs.AI arXiv: https://arxiv.org/abs/2605.27210v1 We adapt Microsoft's QuantumKatas -- a well-established quantum computing curriculum -- from Q# to Qiskit, the most widely-adopted quantum computing framework, and package it with an evaluation framework for systematic LLM assessment. The resulting benchmark comprises 350 tasks across 26 categories, spanning fundamental gates through advanced algorithms (Grover's, Simon's, Deutsch-Jozsa), error correction, key distribution, and quantum games. Each task includes a natural language prompt, canonical solution, and deterministic test verification via classical circuit simulation. By building on the QuantumKatas' proven pedagogical design rather than creating tasks from scratch, we inherit a principled difficulty progression and comprehensive concept coverage while contributing the framework adaptation, evaluation infrastructure, and empirical analysis. We evaluate 16 LLMs across 7 prompting configurations -- a total of 39,200 model runs -- to demonstrate the benchmark's utility. Three key findings emerge: (1) the benchmark effectively differentiates model capabilities, with best-configuration pass rates ranging from 32.3% to 83.1% and a 26.1 pp average gap between frontier and open-source models; (2) models perform well at implementing known algorithms (SimonsAlgorithm 82.1%, BasicGates 81.6%) but struggle with problem encoding (SolveSATWithGrover 34.4%, DistinguishUnitaries 40.0%); and (3) chain-of-thought prompting shows a modestly bimodal effect -- it is the best strategy for three models (two of them explicitly reasoning-tuned per vendor documentation) but degrades performance for the rest, leaving it mid-pack in aggregate (56.3% mean) behind few-shot-5 (57.8%). We release the benchmark, evaluation framework, and baseline results to support research on LLM capabilities in quantum computing. --- ### Learning to Act under Noise: Enhancing Agent Robustness via Noisy Environments Authors: Yuxin Chen, Xiaodong Cai, Junfeng Fang, Zhuowen Han, Yu Wang, Yaorui Shi, Yi Zhang, Qi Gu, Xunliang Cai, Xiang Wang, An Zhang, Tat-Seng Chua Published: 2026-05-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.27209v1 Recent advances in large language models (LLMs) have facilitated the widespread deployment of LLMs as interactive agents capable of reasoning, planning, and tool use. Despite strong performance on existing benchmarks, such agents often exhibit notable degradation when deployed in real-world settings, where environments are inherently stochastic and imperfect. We argue that this discrepancy arises from a fundamental mismatch between idealized training settings and real-world interaction dynamics, where current paradigms rely on carefully curated task instructions and stable, well-controlled environments. To address this gap, we propose NoisyAgent, an agentic training framework that explicitly incorporates environmental imperfections into the agent learning process. We identify two major sources of interaction noise in real-world scenarios: user noise, which captures ambiguity and variability in user interaction, and tool noise, which reflects failures and anomalies in tool execution. We introduce such perturbations into the training pipeline by modifying user interaction patterns and simulating tool execution results within the training environment. To stabilize training while encouraging agents to handle increasingly challenging imperfections, noise is applied to only a subset of rollouts and progressively increased in difficulty as the model adapts to the current noise level. Extensive experiments demonstrate that our approach consistently improves agent robustness under noisy and dynamic environments. Our analysis reveals that training under noise conditions also yields performance gains on idealized benchmarks, suggesting that controlled exposure to environmental noise promotes more generalizable reasoning and decision-making behaviors. Our findings highlight the importance of modeling interaction imperfections for bridging the gap between agent training and real-world deployment. --- ### TWIST: Closed-Loop token Synchronization for Application-Aware Wireless Digital Twins Authors: Sige Liu, Kezhi Wang Published: 2026-05-26 Categories: eess.IV, cs.AI arXiv: https://arxiv.org/abs/2605.27205v1 Wireless digital twins require repeated synchronization between a time-evolving physical scene and its digital counterpart under limited and time-varying communication resources. For perception-centric twins, pixel-domain transmission or uniformly protected bitstreams can be mismatched to the semantic state consumed by twin-side applications. This paper proposes TWIST, a closed-loop token synchronization framework for application-aware wireless digital twins. TWIST represents each physical observation as a token and synchronizes this state over a wireless link, rather than optimizing visual reconstruction. Token positions are grouped by task relevance and protected through mode-conditioned unequal error protection under low-, medium-, and high-synchronization modes. At the twin side, decoding confidence converts unreliable hard token decisions into erasures, which are restored by a completion model before updating the semantic twin state. The recovered state supports traffic-state inference and generates compact feedback statistics, including channel quality, receiver uncertainty, semantic drift, and application priority, for subsequent mode adaptation. Experiments on a dynamic road-scene digital-twin scenario show that TWIST improves traffic-state inference and semantic twin-state synchronization compared with fixed-mode and channel-only adaptation strategies, while reducing the average synchronization cost relative to always-high transmission. --- ### GraphReview: Scientific Paper Evaluation via LLM-based Graph Evidence Expansion Authors: Pujun Zheng, Wanying Ren, Jiacheng Yao, Guoxiu He, Star X. Zhao Published: 2026-05-26 Categories: cs.CL, cs.IR arXiv: https://arxiv.org/abs/2605.27204v2 Scientific paper evaluation often involves not only assessing a manuscript itself, but also relating it to contemporaneous research and prior literature. However, existing LLM-based methods typically model these signals separately and lack a unified mechanism for aggregating review evidence across papers. We propose $\textbf{GraphReview}$, a graph-based LLM framework that formulates paper evaluation as inference-time graph evidence expansion over a semantic paper graph. The graph jointly captures intrinsic quality, synchronic links among contemporaneous papers, and diachronic links to prior work. LLMs are used to estimate node-level quality priors and generate edge-level comparative evidence through pairwise paper comparisons, while Personalized PageRank integrates these structured signals for quality ranking, decision prediction, and review generation. To produce higher-quality graph evidence, we propose reward-induced maximum likelihood objectives for training the LLM backbones. Experiments show that GraphReview consistently outperforms the strongest baseline, achieving average improvements of 29.7% on decision and ranking metrics, including gains of 23.7% in Accuracy and 57.6% in Spearman's $ρ$. It also produces higher-quality review texts and generalizes effectively across time periods and conference venues. --- ### GraphReview: Scientific Paper Evaluation via LLM-Based Graph Message Passing Authors: Pujun Zheng, Wanying Ren, Jiacheng Yao, Guoxiu He, Star X. Zhao Published: 2026-05-26 Categories: cs.CL, cs.IR arXiv: https://arxiv.org/abs/2605.27204v1 Scientific paper evaluation often involves not only assessing a manuscript itself, but also relating it to contemporaneous research and prior literature. However, existing LLM-based methods typically model these signals separately and lack a unified mechanism for propagating review evidence across papers. We propose $\textbf{GraphReview}$, a graph-based LLM framework that formulates paper evaluation as review-signal message passing over a semantic paper graph. The graph jointly captures intrinsic quality, synchronic links among contemporaneous papers, and diachronic links to prior work. LLMs are used to estimate node-level quality priors and generate edge-level comparative evidence through pairwise paper comparisons, while Personalized PageRank integrates review signals for quality ranking, decision prediction, and review generation. To produce higher-quality graph evidence, we propose reward-induced maximum likelihood objectives for training the LLM backbones. Experiments show that GraphReview consistently outperforms the strongest baseline, achieving average improvements of 29.7% on decision and ranking metrics, including gains of 23.7% in Accuracy and 57.6% in Spearman's $ρ$. It also produces higher-quality review texts and generalizes effectively across time periods and conference venues. The code is available at https://github.com/ECNU-Text-Computing/GraphReview. --- ### Generative Animations: A Multi-Model Pipeline for Prompt-Driven Motion Synthesis Authors: Mannat Khurana, Sanyam Jain, Rishav Agarwal Published: 2026-05-26 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.27203v1 Animation elevates digital documents into immersive experiences, yet creating custom motion paths remains cumbersome, requiring designers to manually select presets, plot Bézier points, and configure timing properties. We introduce Generative Animations, a system that transforms natural language prompts into production-ready animations. By chaining Large Language Models (LLMs) for semantic parsing with the Segment Anything Model (SAM) for visual grounding, our pipeline automatically generates motion paths that respect scene geometry, handle depth-based occlusions, and honor 3D perspective transforms. We demonstrate the system through three use cases: contour-following trajectories, orbital animations with z-order awareness, and perspective-aligned motion on transformed objects. --- ### EpiCurveBench: Evaluating VLMs on Epidemic Curve Digitization Authors: Thomas Berkane, Maimuna S. Majumder Published: 2026-05-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.27195v1 Chart-to-data extraction with vision-language models (VLMs) is increasingly evaluated on benchmarks that show diminishing headroom (frontier VLMs exceed 89% on ChartQA) and with metrics that treat extracted points as unordered key-value pairs, ignoring the temporal structure of time series and penalizing small alignment shifts as catastrophic failures. We address both gaps with EpiCurveBench, a benchmark of 1,000 real-world epidemic curve images curated from diverse public-health sources, and EpiCurveSimilarity (ECS), an evaluation metric that aligns predicted and ground-truth series via dynamic programming, tolerating local temporal shifts and gaps while penalizing them proportionally. Evaluating six methods--three frontier closed VLMs, one open VLM, and two specialized chart-extraction systems--we find the strongest model reaches only 52.3% ECS, and that ECS spreads the four general-purpose VLMs over a 25-point range where key-value metrics (RMS, SCRM) compress them into a 5-point band. We further validate ECS against four downstream epidemiological summary statistics, finding that higher ECS predicts smaller errors in total counts, peak timing, and peak magnitude, and higher growth-rate fidelity; across all four, ECS correlates 1.5--3.6 times more strongly than Dynamic Time Warping, which lacks a gap penalty and therefore cannot distinguish a truncated prediction from a temporally faithful one. EpiCurveBench targets a high-impact public-health application--unlocking decades of outbreak data trapped in published figures--but the benchmark and metric apply directly to any structured time-series chart-extraction setting. --- ### Not All Tokens Matter Equally: Dynamic In-context Vector Distillation with Decisive-Token Supervision for Long-form Medical Report Generation Authors: Ning Wu, Rui Liu, Xinkun Lin, Weixing Chen, Jinxi Xiang, Tao Wei, Lina Yao, Mingjie Li Published: 2026-05-26 Categories: cs.CL, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2605.27194v1 Distilling demonstration effects into hidden-space interventions offers a lightweight alternative to full finetuning. However, existing multimodal variants are mostly evaluated on short-form tasks, where outputs end after a few tokens. Extending these methods to long-form generation exposes a fundamental yet underexamined limitation: token-level distillation implicitly treats all output tokens as equally informative, but long-form outputs are dominated by high-frequency template and grammatical tokens, while the tokens that actually determine output quality are sparsely distributed. In medical report generation (MRG), two such decisive tokens stand out: pathology-related tokens that determine diagnostic content, and the end-of-sequence (EOS) event that determines termination. Both receive insufficient supervision under uniform cross-entropy, and autoregressive decoding further compounds the problem by drifting away from teacher-forced trajectories. We propose DIVE, a frozen-backbone distillation framework that addresses long-form report generation through two complementary mechanisms matched to these failures. Decisive-token supervision restores supervision balance by upweighting the cross-entropy contribution of pathology-related tokens and the EOS event, ensuring that content fidelity and termination are learned during training rather than imposed at decoding time. State-conditioned dynamic steering replaces fixed open-loop residuals with hidden-state-dependent adapters, allowing the injected signal to adapt as decoding drifts. Experiments on MIMIC-CXR and CheXpert Plus with two medical VLM backbones show that DIVE consistently ranks among the strongest methods across lexical and clinical-proxy metrics. Our method achieves the best BLEU-4, ROUGE-L, and RadGraph F1 in all dataset--backbone settings, while remaining competitive on coarse label-level CheXbert F1. --- ### Learning When to Think While Listening in Large Audio-Language Models Authors: Zhiyuan Song, Weici Zhao, Yang Xiao, Suhao Yu, Cheng Zhu, Jiatao Gu Published: 2026-05-26 Categories: cs.CL, cs.AI, cs.LG, cs.SD arXiv: https://arxiv.org/abs/2605.27190v1 Recent advances in Large Audio-Language Models (LALMs) have made real-time, streaming spoken interaction increasingly practical. In this setting, reasoning quality and responsiveness are tightly coupled: delaying reasoning until the speech endpoint can improve answer quality but moves deliberation into user-visible response delay, while answering too early risks committing before decisive evidence arrives. We introduce a learnable wait-think-answer control formulation for LALMs. Motivated by the incremental nature of human conversation, the controller decides under partial audio evidence when to wait, when to externalize a compact reasoning update, and when to answer. Using Qwen2.5-Omni-7B as the base model, we construct aligned wait-think-answer traces from spoken reasoning data, train the controller with supervised fine-tuning (SFT), and then apply Decoupled Clip and Dynamic Sampling Policy Optimization (DAPO). The reward combines answer correctness, action validity, update timing, latency synchronization, reasoning quality, and chain consistency, optimizing the complete wait-think-answer trajectory and not the final answer alone. On a six-task synthetic spoken reasoning question answering (SRQA) benchmark, the six-reward DAPO controller improves the row-weighted accuracy from 67.6% to 70.3% while reducing post-endpoint final-think length by 14% under the same Qwen deployment harness. On a 186-item human-recorded Real Audio Bench, a transfer check beyond text-to-speech (TTS)-rendered speech, the controller family remains functional: SFT achieves the strongest accuracy, while the six-reward DAPO controller is the only learned variant whose final-think length falls below the base. These results suggest that a streaming model should learn when to make intermediate reasoning explicit during the audio stream. --- ### Beyond Binary: Speech Representations Across the Cognitive Score Hierarchy Authors: Serli Kopar, Roshan Prakash Rane, Christian Mychajliw, Lydia Federmann, Gerhard Eschweiler, Daniela Berg, Sam Gijsen, Paula Andrea Perez-Toro, Kerstin Ritter Published: 2026-05-26 Categories: cs.CL, cs.LG, cs.SD, eess.AS, q-bio.NC arXiv: https://arxiv.org/abs/2605.27189v1 This study examines the relationship between speech representations and the hierarchical structure of cognitive assessment in mild cognitive impairment. Utilizing 5,754 German neuropsychological assessment recordings, we evaluate six cognitive tasks across three score levels: task, domain, and global levels. We compare hand-crafted acoustic features with self-supervised learning (SSL) embeddings. Results show that although SSL representations generally outperform hand-crafted features at lower levels, this trend reverses for MCI classification. Furthermore, task-specific constraints influence performance: tasks with greater response freedom exhibit performance dilution as hierarchical levels increase, suggesting ``specialist'' representations, whereas the performance of highly structured tasks increases toward higher levels, suggesting ``generalist'' representations. These findings show links between task constraints and assessment hierarchy in automated clinical speech analysis. --- ### MAIGO: Mitigating Lost-in-Conversation with History-Cleaned On-Policy Self-Distillation Authors: Haoyu Zheng, Yun Zhu, Shu Yuan, Shangming Chen, Qing Wang, Wenqiao Zhang, Jun Xiao, Yueting Zhuang Published: 2026-05-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.27186v2 Large language models often solve tasks from a fully specified prompt but degrade when the same requirements unfold over multiple turns, known as the lost-in-conversation (LiC) gap. We trace part of this degradation to self-contamination: intermediate assistant replies enter later context and carry early deviations forward. Motivated by this mechanism, we propose MAIGO, an on-policy self-distillation method that reduces this contamination using history-cleaned references from the model's own policy. For middle turns, MAIGO removes prior assistant replies while preserving the user-visible sharded prefix; for answer turns, it distills from paired full-view references conditioned on the completed user-side dialogue. A reliability weight downweights middle-turn samples that disagree with the clean reference. MAIGO requires no verifier rewards, state labels, or inference-time scaffolding. Under the LiC paired-view protocol with deterministic verifiers, MAIGO improves Qwen2.5-7B-Instruct SHARDED accuracy from 52.8 to 66.1 and the SHARDED/FULL ratio from 66.5% to 84.1%, while keeping FULL accuracy within 2.3 points. These results show that self-contamination is a trainable component of the LiC gap. --- ### MAIGO: Mitigating Lost-in-Conversation with History-Cleaned On-Policy Self-Distillation Authors: Haoyu Zheng, Yun Zhu, Shu Yuan, Shangming Chen, Qing Wang, Wenqiao Zhang, Jun Xiao, Yueting Zhuang Published: 2026-05-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.27186v1 Large language models often solve tasks from a fully specified prompt but degrade when the same requirements unfold over multiple turns, known as the lost-in-conversation (LiC) gap. We trace part of this degradation to self-contamination: intermediate assistant replies enter later context and carry early deviations forward. Motivated by this mechanism, we propose MAIGO, an on-policy self-distillation method that reduces this contamination using history-cleaned references from the model's own policy. For middle turns, MAIGO removes prior assistant replies while preserving the user-visible sharded prefix; for answer turns, it distills from paired full-view references conditioned on the completed user-side dialogue. A reliability weight downweights middle-turn samples that disagree with the clean reference. MAIGO requires no verifier rewards, state labels, or inference-time scaffolding. Under the LiC paired-view protocol with deterministic verifiers, MAIGO improves Qwen2.5-7B-Instruct SHARDED accuracy from 52.8 to 66.1 and the SHARDED/FULL ratio from 66.5% to 84.1%, while keeping FULL accuracy within 2.3 points. These results show that self-contamination is a trainable component of the LiC gap. --- ### FoundObj: Self-supervised Foundation Models as Rewards for Label-free 3D Object Segmentation Authors: Zihui Zhang, Zhixuan Sun, Yafei Yang, Jinxi Li, Jiahao Chen, Bo Yang Published: 2026-05-26 Categories: cs.CV, cs.AI, cs.LG, cs.RO arXiv: https://arxiv.org/abs/2605.27178v1 We address the challenging task of 3D object segmentation in complex scene point clouds without relying on any scene-level human annotations during training. Existing methods are typically constrained to identifying simple objects, primarily due to insufficient object priors in the learning process. In this paper, we present FoundObj, a novel framework featuring a superpoint-based object discovery agent that incrementally merges suitable neighboring superpoints, guided by our innovative semantic and geometric reward modules. These modules synergistically leverage semantic and geometric priors from self-supervised 2D/3D foundation models, providing complementary feedback to the object discovery agent and enabling robust identification of multi-class objects through reinforcement learning. Extensive experiments on diverse benchmarks demonstrate that our approach consistently outperforms existing baselines. Notably, our method exhibits strong generalization in zero-shot and long-tail scenarios, underscoring its potential for scalable, label-free 3D object segmentation. --- ### The Compressive Knowledge Graph Hypothesis: Which Graph Facts Matter for Scientific Hypothesis Generation? Authors: Shashwat Sourav, Viktoriia Baibakova, Sanjay Das, Ran Elgedawy, Maria Mahbub, Emily Herron, Tirthankar Ghosal Published: 2026-05-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.27176v2 Knowledge graphs (KGs) can provide structured scientific context to language models, but it remains unclear which graph facts actually shape the generated hypotheses. We study KG-guided hypothesis generation for battery materials across Mistral-7B, Llama-3.1-70B, and Gemini 2.5 Flash. We perturb local KGs by varying density, ontology richness, topology, and control structure, and evaluate outputs with both provided-graph and fixed-reference metrics. Across models, KG utility is selective and model-dependent: graph context changes outputs, but no-KG outputs also recover substantial graph content from model priors. Compact top-k subgraphs often approximate full-KG behavior, including when claimed-outcome triples are held out. At the same time, compression is not unique to one semantic ranking rule, random and topology-based subsets can also recover much of the signal. These results support a redundancy-aware Compressive KG hypothesis: useful KG signal is often recoverable from compact, scientifically structured subgraphs rather than requiring the full local graph. --- ### An investigation of AI integration in sound designer workflows and experiences Authors: Nelly Garcia, Joshua Reiss Published: 2026-05-26 Categories: cs.SD, cs.AI, cs.CY arXiv: https://arxiv.org/abs/2605.27174v1 Artificial intelligence is increasingly being integrated into professional audio production workflows, yet a gap persists between the tools developers produce and the requirements of practising sound designers. This paper investigates this gap through a mixed-methods study comprising a survey of 76 practitioners and follow-up semi-structured interviews with 20 industry professionals. Results were analysed using descriptive statistical analysis and thematic analysis to identify patterns across both datasets. Five themes emerged from our analysis: Context, Workflow, Potential, Risks, and Right Use. Our work indicates that current AI tools perform adequately in fast-consumption media contexts but lack the narrative sophistication required for high-end sound design (films, immersive experiences etc). Practitioners demonstrate a preference for assistive, task-specific applications, particularly in audio restoration and library management, over end-to-end generative systems. This work contributes to the on-going discussion on the use of AI and AI-enhanced tools in the creative industries. We report on the current status of the field from the point of view of sound designers and creative audio practitioners, and offer a set of recommendation for sound technologist and developers based on our findings to guide the development of more informed AI tools for sound design. --- ### Grounding Text Embeddings in Stakeholder Associations Authors: Jonathan Rystrøm, Sofie Burgos-Thorsen, Zihao Fu, Johan Irving Søltoft, Kenneth C. Enevoldsen, Chris Russell Published: 2026-05-26 Categories: cs.CL, cs.AI, cs.CY arXiv: https://arxiv.org/abs/2605.27168v1 Text embeddings are widely used to analyse large corpora of complex texts. However, it is unclear whether the embeddings capture the same semantic distances as the human experts using them. Ensuring alignment between embedding representations and human intentions is essential for valid analyses. We present the Stakeholder Grounding Exercise, a method for making expert associations explicit and grounding embedding model results in human understanding. In our primary case study on Danish policy issues, we find that neural text embeddings are substantially less reliable than human experts (19-26 pp gap), and that this misalignment propagates to downstream clustering performance (Spearman $ρ=0.9$ between exercise ranking and cluster quality). A secondary study on US Federal AI use cases replicates the gap (16pp) in English, using a digital protocol and a different community of experts -- demonstrating that the gap is not an artefact of a single instrument or domain. The Stakeholder Grounding Exercise offers a practical method for assessing whether embedding models capture the semantic distinctions that matter most to domain experts. --- ### Query Symbolically or Retrieve Semantically? A Dataset and Method for Semi-Structured Question Answering Authors: Mateusz Czyżnikiewicz, Ryszard Tuora, Adam Kozakiewicz, Tomasz Ziętkiewicz, Mateusz Galiński, Michał Godziszewski, Michał Karpowicz, Timothy Hospedales, Cristina Cornelio Published: 2026-05-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.27164v1 Retrieval-Augmented Generation (RAG) systems for question answering typically retrieve evidence by semantic similarity between the query and document chunks. While effective for unstructured text, this approach is less reliable on semi-structured corpora where answering may require exact filtering, aggregation, or exhaustive retrieval over structured attributes across multiple documents. Symbolic approaches support such operations, but they are often brittle on noisy natural-language corpora. We address this gap with DualGraph, a RAG framework that represents documents through two complementary views: a Textual Knowledge Graph for semantic retrieval and a Symbolic Knowledge Graph for symbolic querying over typed subject--predicate--object triples. Building on these two components, we provide multiple strategies for selecting or combining semantic and symbolic evidence.We also introduce SpecsQA, a benchmark from a commercial shopping website with semi-structured product documents and manually curated questions spanning open-ended and specification-oriented retrieval. Experiments show that DualGraph consistently outperforms state-of-the-art dense-retrieval, GraphRAG, symbolic, and table-oriented baselines across question types.Code and data are available at https://github.com/corneliocristina/DualGraphRAG. --- ### The Role of Causal Features in Strategic Classification for Robustness and Alignment Authors: Antonio Gois, Sophia Gunluk, Nir Rosenfeld, Nidhi Hegde, Simon Lacoste-Julien, Dhanya Sridhar Published: 2026-05-26 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2605.27163v1 In strategic classification, an institution (e.g., a bank) anticipates adaptation from users who change their features to increase utility in a classification task (e.g., loan repayment). Since a key challenge is the distribution shift induced by users, we turn to causal models, which have been shown to bound the worst-case out-of-distribution (OOD) risk, and establish several new results that link causality and strategic classification. First, we show that causal classification leads to optimal classification error after any sufficiently large adaptation, when the noise is bounded in a certain way. Second, when these assumptions do not hold, we show OOD cross-entropy risk of optimal classifiers decomposes into an OOD bias term and a term arising from not using all observable features, allowing us to understand when causal classifiers have an advantage. Finally, we show that the use of causal features can allow alignment of long-term incentives between institutions and users, contrasting with previous work that highlights social costs of such approaches. We validate our theory empirically on synthetic data, finding that our results predict behavior in practice. --- ### Formalization of Malagasy conjugation Authors: Joro Ny Aina Ranaivoarison, Eric Laporte, Baholisoa Simone Ralalaoherivony Published: 2026-05-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.27161v1 This paper reports the core linguistic work performed to construct a dictionary-based morphological analyser for Malagasy simple verbs. It uses the Unitex platform and comprised the contruction of an electronic dictionary for Malagasy simple verbs. The data is encoded on the basis of morphological features. The morphological variations of verb stems and their combination with inflectional affixes are formalized in finite-state transducers represented by editable graphs. 78 transducers allow Unitex to generate a dictionary of allomorphs of stems. 271 other transducers are used by the morphological analyser of Unitex to recognize the stem and the affixes in conjugated verbs. The design of the dictionary and transducers prioritizes readability, so that they can be extended and updated by linguists. --- ### Model discovery for dynamical systems with complex-valued product units Authors: Martin Brückmann, Babette Dellen, Uwe Jaekel Published: 2026-05-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.27158v1 Discovering the governing equations of a dynamical system from observed trajectories provides deeper insight into its structure than mere prediction of future states. We present a data-driven approach to model discovery based on complex-valued product-unit networks, in which each unit represents a complex monomial and the network output is a sparse linear combination of such monomials. In contrast to established library-based methods such as SINDy, our approach does not require a predefined set of candidate functions: the relevant monomials, including those with fractional or negative exponents, are learned directly from data. Across four chaotic benchmark systems (Lorenz63, Lorenz84, the Four-Wing attractor, and a fractional variant of Lorenz63), we recover the exact governing equations in 90% of trials for the first three systems, and in 70-90% of trials for the fractional case, using at least 3000 training points. Applied to real-world human-gait accelerometer signals, the model produced stable trajectories with bounded prediction errors, corresponding to an RMSE of approximately 12-14% of the signal amplitude range over a test horizon three times longer than the training interval, demonstrating its potential for high-dimensional systems in which analytic equations are unavailable. --- ### Detecting Is Not Resolving: The Monitoring Control Gap in Retrieval Augmented LLMs Authors: Zhe Yu, Wenpeng Xing, Chen Ye, Xuyang Teng, Bo Yang, Changting Lin, Meng Han Published: 2026-05-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.27157v1 Retrieval-augmented LLMs are deployed for tasks where evidence quality determines action safety, yet evaluation protocols assume that single-turn robustness predicts robustness when evidence accumulates across turns. We show this assumption is fundamentally incorrect. Models exhibit a monitoring-control gap: they readily acknowledge contradictory evidence, yet this awareness fails to constrain their final recommendations - detecting epistemic conflict does not imply resolving it safely. Through a multi-turn document accumulation protocol across four model families (1.5B-32B parameters) and over 50,000 turn-level evaluations, we demonstrate that single-turn diagnostics systematically overestimate RAG safety, that contradiction acknowledgement is uncorrelated with safe resolution, a pattern corroborated by targeted human validation, and that no universal prompt fix exists. Converging mechanism evidence - hidden-state probing, attention analysis, and response-strategy taxonomy - points to action selection as the most plausible locus of the deficit: danger-relevant information is internally represented and receives enhanced attention during unsafe generation, yet fails to constrain output behavior. The gap between what models recognize and what they do must be measured and closed before retrieval-augmented systems can be trusted in high-stakes settings. --- ### LitSeg: Narrative-Aware Document Segmentation for Literary RAG Authors: Ruikang Zhang, Zhanni Chen, Yiqiao Cai, Qi Su Published: 2026-05-26 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.27156v2 Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by incorporating external knowledge, particularly for long-tail domains such as literary works. However, the critical step of document segmentation in RAG remains largely underexplored. Existing strategies typically either ignore semantics or overlook the complicated narrative structures of literary works, often resulting in chunks with fragmented plots and unclear references that hinder retrieval and generation performance. To address this, we propose LitSeg, a novel narrative-theory-guided segmentation framework. By employing multi-stage prompting, LitSeg explicitly extracts valid events, clarifies narrative structures, and locates turning points to inform segmentation. To alleviate the computational overhead of multi-stage inference with large-scale models, we further introduce LitSeg-Lite, a lightweight single-pass chunker fine-tuned on LitSeg-generated data via a two-stage training strategy, distilling the complex process into a single inference pass. Extensive experiments demonstrate that compared to baselines, our methods yield high-quality text chunks that are narratologically coherent and self-contained. This improved segmentation quality enhances retrieval accuracy and context relevance, and boosts downstream QA performance. Ablation studies validate the efficacy of narratological guidance and data distillation, and efficiency analysis shows that LitSeg-Lite matches the teacher at a substantially lower inference cost. --- ### LitSeg: Narrative-Aware Document Segmentation for Literary RAG Authors: Ruikang Zhang, Zhanni Chen, Yiqiao Cai, Qi Su Published: 2026-05-26 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.27156v1 Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by incorporating external knowledge, particularly for long-tail domains such as literary works. However, the critical step of document segmentation in RAG remains largely underexplored. Existing strategies are typically semantically blind and overlook the complicated narrative structures of literary works, often resulting in fragmented plots and unclear references that severely hinder retrieval and generation performance. To address this, we propose LitSeg, a novel narrative-theory-guided segmentation framework. By employing multi-stage prompting, LitSeg explicitly extracts valid events, untangles narrative threads, clarifies narrative structures, and locates turning points to inform segmentation. To alleviate the computational overhead of multi-stage inference with large-scale models, we further introduce LitSeg-Lite, a lightweight single-pass chunker fine-tuned on LitSeg-generated data via a two-stage training strategy, distilling the complex process into a single inference pass. Extensive experiments demonstrate that with structurally independent text chunks, our methods significantly improve retrieval accuracy and context relevance over baselines, ultimately enhancing downstream QA performance, while ablation studies validate the efficacy of narratological guidance and data distillation. --- ### Semantic Robustness Probing via Inpainting: An Interactive Tool for Safety-Critical Object Detection Authors: Nico Steckhan, Krutarth Prajapati, Weija Shao, Silvia Vock Published: 2026-05-26 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.27155v2 Testing object detectors in safety-critical domains requires semantically meaningful probes beyond pixel-level corruptions. We present SemProbe, a tool for semantic robustness probing: users upload deployment images, create masks manually or automatically, select operational design domain-derived factors (or custom prompts), and run diffusion-based controlled inpainting. The system supports batch jobs, parallel seed/workflow variations, and configurable generation parameters. After each output, model inference runs automatically and displays annotated before/after comparisons with performance deltas. All probes are logged as structured artifacts, enabling traceable robustness evidence aligned with safety evaluation workflows. We demonstrate \textsc{SemProbe} on hand detection for dimension saws, targeting factors from insurance-oriented test criteria. --- ### Touch-R1: Reinforcing Touch Reasoning in MLLMs Authors: Yingxin Lai, Yafei Zhou, Fucai Zhu, Siyu Zhu, Weihao Yuan Published: 2026-05-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.27154v1 While rule-based reinforcement learning has recently catalyzed explicit reasoning in multimodal models, tactile reasoning remains largely underexplored. Existing tactile-language models primarily rely on supervised or contrastive objectives, which limits their capacity to ground predictions in physical evidence or rectify misleading visual priors. Tactile reasoning introduces two modality-specific challenges: the ordinal nature of physical attributes (e.g., hardness, roughness) and the cross-sensor distribution shifts inherent in optical tactile hardware. In this work, we introduce TouchReason-1M, a large-scale multimodal dataset comprising over 1M synchronized tactile pairs across four distinct sensors, and TouchReason-Bench, a rigorous framework for evaluating tactile perception and visual-tactile conflict resolution. Building upon these, we propose Touch-R1, a tactile reasoning MLLM based on Qwen2.5-VL-7B. Touch-R1 is trained via a tactile-grounded GRPO objective that combines ordinal-aware accuracy, cross-sensor physical consistency, structured-format control, and an input-side tactile grounding objective. Specifically, the tactile-use reward assigns credit only when authentic tactile inputs yield superior correctness relative to counterfactual controls where the tactile stream is removed, shuffled, or noise-masked. On TouchReason-Bench, Touch-R1-7B outperforms Octopi-13B by 18.4\% and GPT-4o by 24.7\% on average. Its structured reasoning traces reveal emergent behaviors of probing, comparison, and revision, demonstrating that R1-style reasoning can be effectively grounded in physical contact. --- ### Chaos-SSL: An Attention-Based Self-Supervised Learning Framework with Chaotic Transformation for Medical Image Classification Authors: Joao Batista Florindo Published: 2026-05-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.27146v1 Self-Supervised Learning (SSL) has emerged as a powerful paradigm to mitigate the reliance on large, annotated datasets, a common bottleneck in medical image analysis. However, standard SSL methods, which rely on simple geometric and color augmentations, may fail to capture the fine-grained, complex textural details necessary for classifying subtle pathologies. This paper introduces Chaos-SSL, a novel two-stage framework for medical image classification. In the first stage, we propose a new self-supervised pre-training strategy that leverages 1D chaotic maps (Logistic, Tent, and Sine) as a complex, non-linear augmentation for contrastive learning. We hypothesize that these chaotic transformations create ``harder'' and more semantically-rich views, forcing a network to learn robust representations of fine-grained medical textures. In the second stage, we introduce an attention-based fusion model that dynamically combines the specialized features from our Chaos-SSL model with the general-purpose features of a larger, ImageNet-pre-trained model. We validate our method on two public datasets: ISIC 2018 (skin lesions) and APTOS 2019 (diabetic retinopathy). Our results demonstrate that the Chaos-SSL model pre-trained with a Tent map for 30 epochs, followed by attention fusion, achieves performance fully competitive with the state-of-the-art, yielding an accuracy of 0.9261 on ISIC 2018 and 0.8726 on APTOS 2019. This significantly outperforms existing SSL methods, including several recent approaches. ---