Search research

As LLMs move from text completion toward autonomous agents, they remain constrained by the standard chat interface, which lacks private working memory. This raises a fundamental question: can agents reliably perform interactive tasks that depend on hidden state? We define Private State Interactive Tasks (PSITs), which require agents to generate and maintain hidden information while producing public responses consistent with a fixed hidden state. We show theoretically that any agent restricted to the public conversation history cannot both keep the secret unresolved in the transcript and respond consistently with a fixed hidden state in PSITs, yielding an architectural impossibility theorem. To empirically validate this limitation, we introduce a self-consistency testing protocol that evaluates whether agents can maintain a hidden secret across forked dialogue branches. Standard chat-based LLMs and retrieval-based memory baselines fail this test regardless of scale, demonstrating that semantic retrieval does not enable true state maintenance. To address this, we propose a novel architecture incorporating an explicit private working memory; we demonstrate that this mechanism restores consistency with a fixed hidden state, establishing private state as a necessary component for PSIT-capable language agents. Our code is available at https://github.com/chandar-lab/Hangman

Chain-of-Thought (CoT) prompting is widely adopted for mathematical problem solving, including in low-resource languages, yet its behavior under irrelevant context remains underexplored. To systematically study this challenge, we introduce DISTRACTMATH-BN, a Bangla benchmark that augments MGSM and MSVAMP with semantically coherent but computationally irrelevant information. Evaluating seven models ranging from 3B to 12B parameters, we observe substantial performance degradation under distractors: standard models drop by up to 41 points, while reasoning-specialized models decline by 14 to 20 points despite consuming five times more tokens. We propose †DAGGER, which reformulates mathematical problem solving as executable computational graph generation with explicit modeling of distractor nodes. Fine-tuning Gemma-3 models using supervised fine-tuning followed by Group Relative Policy Optimization achieves comparable weighted accuracy on augmented benchmarks while using 89 percent fewer tokens than reasoning models. Importantly, this robustness emerges without explicit training on distractor-augmented examples. Our results suggest that enforcing structured intermediate representations improves robustness and inference efficiency in mathematical reasoning compared to free-form approaches, particularly in noisy, low-resource settings.

Large Language Model (LLM)-based deep research agents perform multi-step reasoning, web exploration, and long-form report generation. In these long-horizon workflows, early deviations from user intent can misdirect research and propagate through planning, search, and synthesis, making timely interaction essential. However, existing benchmarks primarily treat deep research as a static input-output task, overlooking agents' ability to elicit and use user feedback. We introduce IDRBench, a benchmark for evaluating interactive deep research with controlled opportunities for clarification. Within a common workflow and stage-wise interaction budget, IDRBench compares autonomous and interactive trajectories, measuring interaction benefit through changes in task-specific report alignment and interaction cost through turns and tokens. Comprehensive experiments on 100 tasks with seven proprietary and open-weight LLMs show that interaction improves all five alignment measures for every model, yielding an average gain of 6.39 points, while revealing distinct trade-offs among autonomous performance, alignment gain, and communication cost. At the task level, interaction improves performance in 74.4% of cases but degrades it in 19.9%, demonstrating that access to clarification alone does not guarantee better outcomes: success depends on what agents ask and how effectively they incorporate the resulting feedback.

The real-world information sources are inherently multilingual, which naturally raises a question about whether language models can synthesize information across languages. In this paper, we introduce a simple two-hop question answering setting, where answering a question requires making inferences over two multilingual documents. We find that language models are more sensitive to language variation in answer-span documents than in those providing bridging information, despite the equal importance of both documents for answering a question. Under a step-by-step sub-question evaluation, we further show that in up to 33% of multilingual cases, models fail to infer the bridging information in the first step yet still answer the overall question correctly. This indicates that reasoning in language models, especially in multilingual settings, does not follow a faithful step-by-step decomposition. Subsequently, we show that the absence of reasoning decomposition leads to around 18% composition failure, where both sub-questions are answered correctly but fail for the final two-hop questions. To mitigate this, we propose a simple three-stage SUBQ prompting method to guide the multi-step reasoning with sub-questions, which boosts accuracy from 10.1% to 66.5%.

Building NLP systems for subjective tasks requires one to ensure their alignment to contrasting human values. We propose the MultiCalibrated Subjective Task Learner framework (MC-STL), which clusters annotations into identifiable human value clusters by three approaches (similarity of annotator rationales, expert-value taxonomies or rater's sociocultural descriptors) and calibrates predictions for each value cluster by learning cluster-specific embeddings. We demonstrate MC-STL on several subjective learning settings, including ordinal, binary, and preference learning predictions, and evaluate it on multiple datasets covering toxic chatbot conversations, offensive social media posts, and human preference alignment. The results show that MC-STL consistently outperforms the baselines that ignore the latent value structure of the annotations, delivering gains in discrimination, value-specific calibration, and disagreement-aware metrics.

Semantic Multi-Object Tracking (SMOT) is evolving from purely geometric localization toward comprehensive video understanding. However, existing paradigms predominantly rely on closed-set interaction tags and fragmented perception pipelines, creating a bottleneck that prevents the full utilization of Multi-modal Large Language Models (MLLMs) for dynamic scenes. In this paper, we elevate SMOT from rigid classification to an open-ended generative reasoning task. To support this paradigm shift, we introduce Grand-SMOT, a large-scale benchmark featuring high-density, dual-stream narratives. This dataset explicitly decouples micro-level individual dynamics from macro-level environmental contexts, directly resolving the semantic scarcity of prior tracking datasets. Furthermore, we propose LLMTrack, a unified MLLM-driven framework for dynamic SMOT. Guided by a verifiable ``\textit{Macro-Understanding-First}'' mechanism, LLMTrack employs a Spatio-Temporal Fusion Module to compress discrete geometric trajectories into continuous semantic tokens, effectively suppressing temporal hallucinations in long-sequence tracking. Extensive experiments, utilizing a novel decoupled evaluation protocol, validate that LLMTrack achieves state-of-the-art geometric tracking robustness while delivering a qualitative leap in generative semantic reasoning. The code and datasets are publicly available at https://github.com/liaopan-lp/LLMTrack-GrandSMOT.

We present Exposía, the first public dataset that connects writing and feedback in higher education, enabling research on educationally grounded computational approaches to teaching and evaluating academic writing. Exposía includes student research project proposals and peer and instructor feedback consisting of comments and free-text reviews. The dataset was collected in the "Introduction to Scientific Work" course of the Computer Science. Exposía reflects the multi-stage nature of the academic writing process that includes drafting, receiving feedback, and revising the writing based on the feedback received. Both the project proposals and peer feedback are accompanied by human assessment scores based on a fine-grained, pedagogically-grounded schema for writing and feedback assessment that we develop. We use Exposía to benchmark state-of-the-art large language models (LLMs) on two tasks: automated scoring of (1) the proposals and (2) the student reviews. We find that the two tasks are best served by different LLMs. Furthermore, closed-source models consistently outperform open-weight models, motivating further research on improving the performance of open-weight models preferred in classroom settings. Finally, we establish that a prompting strategy that scores multiple aspects of the writing together is the most effective, paving the way for more effective classroom deployment of modern LLMs.

Persistence models often conflate propagation, survival, and cross-scale loss. The Replicator-Optimization Mechanism (ROM) is a replicator-mutator template separating baseline weight, bounded survival, and a transfer kernel at a declared scale. Its equation conserves mass but guarantees neither invariance, convergence, a potential, nor a preferred scale. For finite static density-independent continuous time, an irreducible weighted kernel yields a unique positive Perron-Frobenius composition; discrete-time power convergence needs primitivity. The componentwise ranking proved here is guaranteed under exact uniform-residual transfer. Strong lumpability gives universal first-order transfer closure, and blockwise effective fitness gives an exact quotient. An institutional instantiation uses normalized stakes, signed preference-decision alignment, information loss, and descriptive effective voice. It specifies conditional survival, not legitimacy or normative authority. A companion mixed-motive MARL battery reports exploratory evidence against the implemented proxy ratio in its environment: a positive signed target-coordinate-correlation effect survives held-out evaluation under shared-state contention, while a reduced feasible-centred frozen-policy crossing reverses the predicted correlation-noise interaction. The treatment varies ideal-point correlation inside a fixed reward family, not objective- or reward-function alignment. Lean checks mapped algebraic identities and scalar monotonicities, not the stationary theorem, empirical mapping, or normative bridge. ROM is an assumptions ledger and model-construction discipline, not a cross-substrate law.

We present a systematic review of 337 articles evaluating the syntactic abilities of Transformer-based language models (TLMs), reporting on over 3,000 datapoints spanning a wide range of syntactic phenomena, languages, models, and methods. We take the data to collectively show that TLMs encode a non-trivial amount of syntactic knowledge. Behavioral evidence shows strong performance on formal syntactic phenomena, but weaker and more variable performance on phenomena at the syntax-semantics interface. Performance is also consistently lower for languages with less digital support. Probing and mechanistic studies further support the presence of syntactic knowledge in TLMs. Yet, because most work remains observational and methodologically heterogeneous, insight into the detailed computational mechanisms underlying syntactic processing remains limited. At the same time, the literature remains heavily concentrated on English and BERT-like models. We discuss the implications of our results and provide recommendations for future research.

Despite strong performance on many tasks, large language models (LLMs) show limited ability in historical and cultural reasoning, particularly in non-English contexts such as Chinese history. Taxonomic structures offer an effective mechanism to organize historical knowledge and improve understanding. However, manual taxonomy construction is costly and difficult to scale. Therefore, we propose \textbf{CHisAgent}, a multi-agent LLM framework for historical taxonomy construction in ancient Chinese contexts. CHisAgent decomposes taxonomy construction into three role-specialized stages: a bottom-up \textit{Inducer} that derives an initial hierarchy from raw historical corpora, a top-down \textit{Expander} that introduces missing intermediate concepts using LLM world knowledge, and an evidence-guided \textit{Enricher} that integrates external structured historical resources to ensure faithfulness. Using the \textit{Twenty-Four Histories}, we construct a large-scale, domain-aware event taxonomy covering politics, military, diplomacy, and social life in ancient China. Extensive reference-free and reference-based evaluations demonstrate improved structural coherence and coverage, while further analysis shows that the resulting taxonomy supports cross-cultural alignment.

Accurate blood glucose forecasting using continuous glucose monitoring (CGM) data can support the early prediction of dysglycemic risk. However, current neural-network-based forecasting models treat CGM data as a purely numerical sequence without integrating the contextual information contained in CGM signal morphology. Recently, large language models (LLMs) have shown promise for time-series forecasting, yet their role as agentic context extractors in diabetes care remains largely unexplored. In this study, we bridge glucose forecasting and LLM-based contextualization by developing GlyRAG, a context-aware, retrieval-augmented forecasting framework that uses an LLM as a contextualization agent to summarize glucose morphology directly from a timed CGM window. The generated CGM-only narrative is embedded and fused with patch-based glucose representations, while a retrieval module incorporates similar historical training episodes through cross-attention. We evaluate GlyRAG on the OhioT1DM and AZT1D datasets for 5-, 30-, and 60-minute forecasting horizons. Compared with strong CGM-only baselines, GPT-4 GlyRAG significantly improves long-horizon root mean square error (RMSE) over PatchTST on both datasets. For example, RMSE decreases from 13.8 to 10.6 at 30 minutes and from 23.1 to 20.2 at 60 minutes on OhioT1DM. LLaMA 3.1 produces smaller but significant long-horizon gains, suggesting that the contextualization pipeline is not limited to GPT-4. Clinical error-grid analyses further show that approximately 85 percent of predictions fall within the clinically acceptable Clarke Error Grid Zones A and B. These results suggest that CGM-derived linguistic context and case-based retrieval can improve long-horizon glucose forecasting without requiring additional sensing modalities.

Automatic speech editing aims to modify spoken content based on textual instructions, yet traditional cascade systems rely on explicit temporal alignment and complex preprocessing. To address these limitations, we propose CosyEdit, an end-to-end speech editing model adapted from CosyVoice through task-specific post-training and a complementary training paradigm, which internalizes text--speech alignment while ensuring high consistency between the speech before and after editing. Trained on only 250 hours of supervised data from our curated GigaEdit dataset, our 400M-parameter model achieves reliable speech editing performance. Extensive evaluations show that CosyEdit not only outperforms several billion-parameter language model baselines but also approaches state-of-the-art cascade systems. These results show that robust and efficient speech editing can be unlocked from a zero-shot TTS model through post-training, offering a cost-effective end-to-end solution for high-quality speech editing. Code and audio samples are available at https://cjy1018.github.io/CosyEditDemoPage/.

Nighttime color constancy still remains a challenging problem in computational photography due to low-light noise and complex illumination conditions. We present RL-AWB, a novel framework combining statistical methods with deep reinforcement learning for nighttime white balance. Our method begins with a statistical algorithm tailored for nighttime scenes, integrating salient gray pixel detection with novel illuminant estimation. Building on this foundation, we develop the first deep reinforcement learning approach for color constancy that leverages the statistical algorithm as its core, mimicking professional AWB tuning experts by dynamically determining image-specific parameters at inference time, without requiring ground-truth illuminants or reference images. To further facilitate cross-sensor evaluation, we introduce the first multi-sensor nighttime dataset. Experiment results demonstrate that our method achieves strong generalization capability across low-light and well-illuminated images. Project page: https://ntuneillee.github.io/research/rl-awb/

While Large Language Model (LLM) role-playing agents have advanced rapidly, it remains unclear which profile elements genuinely drive role-playing quality. To bridge this gap, we introduce a systematic diagnostic framework that disentangles the impact of character profiles along three axes: Familiarity (Known vs. Unknown), Structure (Structured vs. Unstructured), and Disposition (Moral vs. Immoral). Utilizing a unified hierarchical schema (5 dimensions, 28 fields), we construct a controlled dataset of 211 personas and evaluate five LLMs on both single- and multi-turn interactions. Our results reveal a striking asymmetry: \textbf{Familiarity} and \textbf{Structure} show negligible impact, while \textbf{Disposition} produces large, consistent performance degradation for immoral characters across all conditions. Further analyses suggest that the Moral--Immoral gap is amplified by post-SFT alignment, and that this degradation varies substantially across profile attributes. To mitigate this bottleneck, we propose Field-Aware Contrastive Decoding (FACD), a training-free strategy that amplifies suppressed disposition-sensitive signals, significantly closing the performance gap without sacrificing moral-character performance.

Understanding biomedical experiments provides a foundation for downstream tasks, e.g., laboratory automation, and facilitates effective cross-disciplinary communication. Two challenges, High Information Density (HID) and Multi-Step Reasoning (MSR), pose unique difficulties for precise experimental understanding. Extracting structured knowledge, e.g., Knowledge Graphs (KGs), is an effective approach to address the HID and MSR. However, existing biomedical datasets for structured knowledge information extraction are limited to a general or coarse-grained level, hindering fine-grained experimental understanding. To address this gap, we introduce Biomedical Protocol Information Extraction Dataset (BioPIE), a dataset providing procedure-centric KGs that capture entities, actions, and relations at a scale sufficient for reasoning across biomedical protocols. We evaluate information extraction methods on BioPIE and implement a question answering system leveraging the dataset for validation, demonstrating improved understanding performance on test sets as well as on the HID and MSR question sets.

High-quality methodology figures are central to scientific communication, yet they remain difficult and time-consuming to create. Such figures must distill a method's components and information flow into a clear, revisable diagram as the paper evolves. Existing methodology diagram automation systems typically face a trade-off between editability and visual quality: TikZ- or SVG-based methods produce editable structured outputs but often lack the richness of human-designed figures, while image-generation models produce polished raster outputs that are difficult to revise. We introduce SciFig, an end-to-end multi-agent framework for generating visually rich and fully editable methodology figures from scientific text. SciFig decomposes figure generation into planning, layout synthesis, component rendering, and iterative refinement, producing XML figures that can be edited in standard diagramming tools and refined through human or VLM feedback. We also introduce SciFig-Bench, a human-verified benchmark of 435 author-drawn methodology figures from 37 arXiv domains and 15 top-tier AI/ML venues, and SciFig-Eval, a four-axis evaluation protocol for measuring figure quality. Across seven single-agent and agentic baselines, SciFig achieves the best performance on all four SciFig-Eval axes and generates editable figures in about 10 minutes on average. Qualitative examples further show that SciFig can generalize beyond methodology figures to teaser diagrams and statistical plots. Dataset and code are available at: https://shramanpramanick.github.io/SciFig/.

Reliable long-term decoding of gestures from surface electromyography (EMG) is hindered by signal drift caused by electrode displacement, muscle fatigue, and/or posture changes. Although modern models achieve high intra-session accuracy, their performance often degrades substantially across recording sessions. Existing approaches to mitigate this problem typically rely on large training datasets or computationally intensive pipelines that are unsuitable for energy-efficient wearable devices. We propose a lightweight test-time adaptation framework for EMG decoding. The framework includes three complementary adaptation strategies: (i) causal adaptive batch normalization for online statistical alignment, (ii) Gaussian Mixture Model alignment with experience replay to mitigate forgetting, and (iii) meta-learning for rapid few-shot calibration. We evaluate these methods on the multi-session NinaPro DB6 dataset. All approaches substantially improve inter-session robustness relative to a non-adaptive baseline while maintaining low computational overhead. Replay-regularized statistical alignment provides the most stable adaptation under limited data, while meta-learning achieves the highest accuracy when sparse calibration labels are available. Overall, our self-supervised test-time adaptation methods reach up to 82% inter-session accuracy, significantly improving upon prior approaches while maintaining resource-efficient operation. These results demonstrate that lightweight test-time adaptation can enable robust, long-term EMG decoding for wearable or prosthetic applications.

GUI agents that interact with graphical interfaces on behalf of users represent a promising direction for practical AI assistants. However, training such agents is hindered by the scarcity of suitable environments. We present InfiniteWeb, a system that automatically generates functional web environments at scale for GUI agent training. While LLMs perform well on generating a single webpage, building a realistic and functional website with many interconnected pages faces challenges. We address these challenges through unified specification, task-centric test-driven development, and a combination of website seed with reference design image to ensure diversity. Our system also generates verifiable task evaluators enabling dense reward signals for reinforcement learning. Experiments show that InfiniteWeb surpasses commercial coding agents at realistic website construction, and GUI agents trained on our generated environments achieve significant performance improvements on OSWorld and Online-Mind2Web, demonstrating the effectiveness of proposed system.

Mobile GUI agents have shown strong potential in real-world automation and practical applications. However, most existing agents remain reactive, making decisions mainly from current screen, which limits their performance on long-horizon tasks. Building a world model from repeated interactions enables forecasting action outcomes and supports better decision making for mobile GUI agents. This is challenging because the model must predict post-action states with spatial awareness while remaining efficient enough for practical deployment. In this paper, we propose MobileDreamer, an efficient world-model-based lookahead framework to equip the GUI agents based on the future imagination provided by the world model. It consists of textual sketch world model and rollout imagination for GUI agent. Textual sketch world model forecasts post-action states through a learning process to transform digital images into key task-related sketches, and designs a novel order-invariant learning strategy to preserve the spatial information of GUI elements. The rollout imagination strategy for GUI agent optimizes the action-selection process by leveraging the prediction capability of world model. Experiments on Android World show that MobileDreamer achieves state-of-the-art performance and improves task success by 5.25%. World model evaluations further verify that our textual sketch modeling accurately forecasts key GUI elements.

Decarbonization of isolated or off-grid energy systems through phase-in of large shares of intermittent solar or wind generation requires co-installation of energy storage or continued use of existing fossil dispatchable power sources to balance supply and demand. The effective CO2 emission reduction depends on the relative capacity of the energy storage and renewable sources, the stochasticity of the renewable generation, and the control of the isolated energy system. While the operation of the energy storage and dispatchable sources impacts the optimal sizing of the system, it is challenging to account for the effect of finite-horizon optimal control at the stage of system sizing. In this work, we present a flexible and computationally efficient sizing framework for energy storage and renewable capacity in isolated energy systems, accounting for uncertainty in the renewable generation and the optimal control. We implement an imitation learning approach to stochastic neural model predictive control (MPC) which allows us to relate the battery storage and wind peak capacities to the emissions reduction and investment costs while accounting for finite horizon, optimal control without solving an infeasible number of optimization problems. We evaluate the proposed sizing framework on a case study of an offshore energy system with a gas turbine, a wind farm and a battery energy storage system (BESS). In this case, we find a nonlinear, nontrivial relationship between the investment costs and the reduction in gas usage for different wind and BESS capacities.

Physics-Informed Neural Networks (PINNs) provide a framework for integrating physical laws with data. However, their application to Prognostics and Health Management (PHM) remains constrained by the limited uncertainty quantification (UQ) capabilities. Most existing PINN-based prognostics approaches are deterministic or account only for epistemic uncertainty, limiting their suitability for risk-aware decision-making. This work introduces a heteroscedastic Bayesian Physics-Informed Neural Network (B-PINN) framework that jointly models epistemic and aleatoric uncertainty, yielding full predictive posteriors for spatiotemporal insulation material ageing estimation. The approach integrates Bayesian Neural Networks (BNNs) with physics-based residual enforcement and prior distributions, enabling probabilistic inference within a physics-informed learning architecture. The framework is evaluated on transformer insulation ageing application, validated with a finite-element thermal model and field measurements from a solar power plant, and benchmarked against deterministic PINNs, dropout-based PINNs (d-PINNs), and alternative B-PINN variants. Results show that the proposed B-PINN provides improved predictive accuracy and better-calibrated uncertainty estimates than competing approaches. A systematic sensitivity study further analyzes the impact of boundary-condition, initial-condition, and residual sampling strategies on accuracy, calibration, and generalization, and the influence of measurement noise on aleatoric uncertainty. Overall, the findings highlight the capability of Bayesian physics-informed learning to support uncertainty-aware prognostics and informed decision-making in transformer asset management by tracking aleatoric and epistemic sources of uncertainty.

Reliable epidemiological reasoning requires synthesizing study evidence to infer disease burden, transmission dynamics, and intervention effects at the population level. Existing medical question answering benchmarks primarily emphasize clinical knowledge or patient-level reasoning, yet few systematically evaluate evidence-grounded epidemiological inference. We present EpiQAL, to our knowledge the first diagnostic benchmark for epidemiological question answering over research literature, comprising three subsets built from open-access articles across diverse diseases. The three subsets progressively test factual recall, multi-step inference, and conclusion reconstruction under incomplete information, and are constructed through a quality-controlled pipeline combining taxonomy guidance, multi-model verification, and difficulty screening. Experiments on fifteen models spanning open-source and proprietary systems reveal that current LLMs show limited performance on epidemiological reasoning, with multi-step inference posing the greatest challenge. Model rankings shift across subsets, and scale alone does not predict success. Chain-of-Thought prompting benefits multi-step inference but yields mixed results elsewhere. EpiQAL provides fine-grained diagnostic signals for evidence-grounding, inferential reasoning, and conclusion reconstruction.

Background and Objective: Precise and scalable instance segmentation of cell nuclei is a fundamental prerequisite for computational pathology, yet gigapixel whole-slide images (WSIs) pose significant computational challenges. While patch-based processing is standard during training, existing methods are often limited to small tile sizes during inference due to architectural bottlenecks or reliance on computationally expensive post-processing for instance separation. We introduce a faster, scalable, and end-to-end framework capable of processing large-scale image tiles while accurately modeling biologically realistic overlapping nuclei. Methods: We propose LSP-DETR (Local Star Polygon DEtection TRansformer). The model represents nuclei as star-convex polygons and employs a lightweight transformer with linear complexity, enabling the processing of high-resolution images in a single forward pass. A novel radial distance loss accommodates annotation uncertainty, allowing the segmentation of overlapping nuclei to emerge naturally without explicit overlap labels. Results: LSP-DETR achieves state-of-the-art efficiency, with an inference time of 0.45 s/mm^2, a 3.2x speedup over StarDist, the next-fastest method. On PanNuke, the model achieves competitive accuracy (67.5 bPQ), while yielding an F$_1$-score of 0.964 in polygon overlap when evaluated against consensus annotations from two expert pathologists. Furthermore, it outperforms larger models such as LKCell in generalization robustness, reaching an F1-score of 85.0 on MoNuSeg. Conclusions: LSP-DETR bridges the gap between high-fidelity segmentation and practical clinical requirements by eliminating heuristic post-processing. By providing a scalable, linear-complexity solution that naturally handles overlaps between nuclei, this framework sets a new direction for efficient high-throughput WSI analysis in digital pathology.

Humor preferences vary widely across individuals and cultures, complicating the evaluation of humor using large language models (LLMs). In this study, we model heterogeneity in humor preferences in Oogiri, a Japanese creative response game, by clustering users with voting logs and estimating cluster-specific weights over interpretable preference factors using Bradley-Terry-Luce models. We elicit preference judgments from LLMs by prompting them to select the funnier response and found that user clusters exhibit distinct preference patterns and that the LLM results can resemble those of particular clusters. Finally, we demonstrate that, by persona prompting, LLM preferences can be directed toward a specific cluster. The scripts for data collection and analysis are publicly available to support reproducibility.

Recent approaches in personalized reward modeling have primarily focused on leveraging user interaction history to align model judgments with individual preferences. However, existing approaches largely treat user context as a static or implicit conditioning signal, failing to capture the dynamic and multi-faceted nature of human judgment. In this paper, we propose P-Check, a novel personalized reward modeling framework, designed to train a plug-and-play checklist generator that synthesizes dynamic evaluation criteria for guiding the reward prediction. To better align these checklists with personalized nuances, we introduce Preference-Contrastive Criterion Weighting, a training strategy that assigns saliency scores to criteria based on their discriminative power for personalized judgment. We conduct extensive experiments and demonstrate that P-Check not only improves reward accuracy but also enhances downstream personalized generation, and remains robust in OOD scenarios.

Aligning language models to qualitative behavioral traits, such as human-likeness, remains difficult because they are hard to define, measure, and optimize. As a result, improvements in human-like behavior are largely driven by scale or broad supervised training, rather than targeted alignment. We introduce Human Aligning LLMs (HAL), a framework for aligning language models to conversational human-likeness using an interpretable, data-driven reward. HAL derives explicit conversational traits from contrastive dialogue data, combines them into a compact scalar score, and uses this score as a transparent reward signal for alignment with standard preference optimization methods. Using this approach, we align models of varying sizes without affecting their overall performance. In large-scale Chatbot Arena-style human evaluations, a model aligned with HAL is more frequently perceived as human-like in conversation. Because HAL operates over explicit, interpretable traits, it enables inspection of alignment behavior and diagnosis of unintended effects. More broadly, HAL demonstrates how soft, qualitative properties of language--previously outside the scope for alignment--can be made measurable and aligned in an interpretable and explainable way.

State space models (SSMs) like Mamba offer efficient alternatives to Transformer-based language models, with linear time complexity. Yet, their adversarial robustness remains critically unexplored. This paper studies the phenomenon whereby specific short input phrases induce a partial amnesia effect in such models, by irreversibly overwriting information in their hidden states, referred to as a Hidden State Poisoning Attack (HiSPA). Our benchmark RoBench-25 allows evaluating a model's information retrieval capabilities when subject to HiSPAs, and confirms the vulnerability of SSMs against such attacks. Even the recent Jamba-1.7-Mini SSM--Transformer (a 52B hybrid model) collapses on RoBench-25 under some HiSPA triggers, whereas pure Transformers do not. We also observe that HiSPA triggers significantly weaken the Jamba model on the popular Open-Prompt-Injections benchmark, unlike pure Transformers. We further show that the theoretical and empirical findings extend to Mamba-2, and also analyse a Mamba-2-based hybrid (Nemotron-3-Nano). Finally, our interpretability study reveals patterns in Mamba's hidden layers during HiSPAs that could be used to build a HiSPA mitigation system. The full code and data to reproduce the experiments can be found at https://github.com/TortueSagace/hispa.

Spatial computation in geographic systems increasingly requires query-conditioned, local, interpretable aggregation under metric constraints. Many classical approaches rely on global summation and treat approximation as an implementation concern, limiting interpretability and scalability at large scales. We propose the Adaptive Density Field (ADF), a geometric attention framework that formulates spatial aggregation as a query-conditioned, metric-induced attention operator in continuous space. Given a set of labelled spatial points with associated scalar scores, ADF defines a continuous intensity field over space. For a given query location, the field value is obtained via a local adaptive Gaussian kernel mixture centered on the query's nearest neighbors, where kernel bandwidths are modulated by point-specific scores to evaluate local aggregated influence. Additionally, approximate nearest-neighbor search is introduced, enabling scalable execution while preserving locality. The proposed ADF bridges concepts from adaptive kernel methods, classical GIS methods, and attention mechanisms by reinterpreting spatial influence as geometry-embedded attention, grounded in physical distance rather than learned latent projections. The proposed framework is formulation-level rather than algorithm-specific, allowing flexible kernel choices, score-to-bandwidth mappings, and approximation parameters. This approach provides a unifying perspective on spatial influence modeling that emphasizes structure, scalability, and geometric interpretability, with relevance to geographic information systems and spatial machine learning.

Speaker-Attributed, Time-Stamped Transcription (SATS) aims to transcribe what is said and to precisely determine the timing of each speaker, which is particularly valuable for meeting transcription. Existing SATS systems rarely adopt an end-to-end formulation and are further constrained by limited context windows, weak long-range speaker memory, and the inability to output timestamps. To address these limitations, we present MOSS Transcribe Diarize, a unified multimodal large language model that jointly performs Speaker-Attributed, Time-Stamped Transcription in an end-to-end paradigm. Trained on extensive real wild data and equipped with a 128k context window for up to 90-minute inputs, MOSS Transcribe Diarize scales well and generalizes robustly. Across comprehensive evaluations, it outperforms state-of-the-art commercial systems on multiple public and in-house benchmarks.

We propose a reinforcement learning (RL) framework for \xy{responsive} precision tuning for linear solvers, which can be extended to general algorithms. The framework is formulated as a contextual bandit problem and solved using incremental action-value estimation with a discretized state space to select optimal precision configurations for computational steps, \xy{retaining} precision and computational efficiency. To verify its effectiveness, we apply the framework to iterative refinement for solving linear systems $Ax = b$. In this application, our approach dynamically chooses precisions based on calculated features from the system while maintaining acceptable accuracy and convergence. In detail, an action-value estimator takes discretized features (e.g., approximate condition number and matrix norm) as input and outputs estimated action values, from which a policy selects the actions (chosen precision configurations for specific steps), optimized via an $ε$-greedy strategy to maximize a multi-objective reward to balance accuracy and computational cost. Empirical results demonstrate effective precision selection, \xy{increasing the use of lower-precision arithmetic} while maintaining accuracy comparable to double-precision baselines. \xy{We further evaluate the learned policies in a compiled CPU GMRES-IR implementation using FP16, FP32, and FP64 arithmetic for solver-level native validation.} The framework generalizes to diverse out-of-sample data and provides insights into applying RL precision selection to other numerical algorithms, advancing mixed-precision numerical methods in scientific computing. To the best of our knowledge, this is the first work on precision autotuning with RL with verification on unseen datasets.