Search arXiv⌕ Search

arXiv subjects

Bo Han

Publications and source records attributed to Bo Han.

At least 37 records · Page 2Linked to original sources
arXiv2026

SquidAgent: Parallelize Wisely, Coordinate Efficiently

LLM-based agents solve complex multi-step tasks, but sequential execution incurs substantial latency. In principle, parallelizing work across multiple agents should yield near-linear speedups. Yet existing parallel multi-agent systems often run slower than a single-agent baseline. We attribute this gap to two hidden costs that parallel execution incurs but a serial agent avoids. First, there is a re-exploration cost: redundant effort spent by parallel workers reconstructing context that the orchestrator already possesses, such as prior decisions, that would otherwise be inherited implicitly in a serial execution. Second, there is an alignment cost: the overhead required to reconcile inconsistencies across independently generated outputs. We thus derive a principled decision criterion: a layer should be parallelized only when its critical-path cost, plus re-exploration and alignment overheads, is lower than the corresponding serial cost. While this criterion is naturally expressed in wall-clock time, we observe that LLMs are poorly calibrated when asked to estimate task duration. To address this, we instead measure cost in predicted output tokens, which we empirically find LLMs can estimate substantially more reliably than wall-clock time. Building on this token-based criterion, we propose SquidAgent. It estimates all token budgets in a single planning step, forks each worker directly from the orchestrator's session to eliminate re-exploration cost, and replaces post-hoc reconciliation with a pre-generated shared convention block that converts alignment into a bounded upfront cost. A deterministic scheduler then applies the criterion layer by layer. Empirically, SquidAgent achieves a 2.2$\times$ mean throughput improvement and a 2.6$\times$ mean wall-time speedup over Claude Code, and a 2.0$\times$ throughput improvement over the strongest multi-agent baseline.

cs.AI↗
arXiv2026

Test-time Calibration Learning for Large Language Model Reasoning

Reliable large language models (LLMs) must not only produce accurate answers but also express confidence that faithfully reflects their probability of being correct. Such calibration is essential for identifying uncertain predictions and supporting reliable decision-making in real-world deployment. Recent studies incorporate calibration learning into reinforcement learning (RL), jointly optimizing answer correctness and verbalized confidence using ground-truth correctness supervision. However, their reliance on labeled data limits their applicability in practical test-time settings, where ground-truth labels are unavailable and calibration may need to adapt to newly encountered target tasks. To address this challenge, we propose Test-Time Calibration Learning (TTCL), a label-free framework that jointly adapts reasoning accuracy and verbalized confidence directly on unlabeled target-task data. Specifically, TTCL derives self-supervision signals for both correctness and calibration from multiple model-generated responses, enabling calibration learning at test time without ground-truth labels. Theoretical analysis further establishes TTCL as a bounded surrogate for the ideal calibration objective. Extensive experiments on mathematical reasoning and factual question answering demonstrate that TTCL consistently improves both accuracy and calibration across diverse models and tasks. On base models, TTCL achieves an average relative accuracy improvement of +40.13% and an ECE reduction of +70.80% across eight benchmarks. Moreover, TTCL can further improve both accuracy and calibration for already calibrated models under domain shift, particularly when source-domain calibration transfers poorly to target tasks. In the math-to-factQA setting, TTCL achieves an average relative accuracy gain of +20.35% and reduces ECE by +53.83%. The source code is released at https://github.com/tmlr-group/TTCL.

cs.LG↗
arXiv2026

Homological Thresholds in Randomly Monitored Quantum Error-Correcting Codes

Random projective measurements are an effective tool to model the interlinked phenomenology of monitoring, inferring, and decoding highly entangled states such as quantum error-correcting codes. At the same time, the study of such monitored dynamics has opened a new arena for critical phenomena, renewing longstanding questions in disordered statistical mechanics, including percolation, random-bond models, and Nishimori physics. Here we investigate the robustness of stabilizer quantum error-correcting codes under a single round of independently sampled single-qubit Pauli measurements and demonstrate that their threshold behavior falls into two broad categories - geometric and homological percolation transitions, with the toric code and color code being the principal examples for each class. We establish the distinct character of these two threshold theories by analyzing the logical-support criteria of these codes and by providing numerical estimates of their respective critical exponents obtained from optimized stabilizer simulations of large-scale systems with over one million qubits. Our framework extends to non-CSS, subsystem, higher-dimensional, and Floquet stabilizer codes and allows us to determine their respective threshold phase diagrams. We exemplify how the latter provide quantitative and qualitative guidance for the computationally more demanding problems of tunable learning and maximum-likelihood decoding. As a practical application of our framework, we discuss the robustness of the gross code.

quant-ph↗
arXiv2026

Learning Reliable GUI Agents under Imperfect Priors

GUI agents built on large language and vision-language models still struggle on unseen applications and complex multi-step tasks, as completing real GUI tasks depends on app-specific, temporally volatile operational knowledge that is scarce in pretraining corpora. Retrieval-augmented execution offers a natural remedy but faces two coupled bottlenecks: knowledge at scale is hard to acquire, and self-collected priors inevitably drift from the live environment due to version updates, promotions, ads, A/B tests, and personalization. We therefore argue that GUI agents should not pursue perfect knowledge but learn to act correctly under imperfect priors, and propose our framework that couples knowledge acquisition with noise-robust utilization: a structured exploration strategy traverses interactive elements, builds a UI state-transition graph, and synthesizes (task, trajectory) pairs via a VLM without human annotation; a noise-aware training strategy, grounded in a taxonomy of real GUI drift patterns, injects five types of realistic errors into self-explored trajectories to teach the agent to assess prior reliability before acting. Experiments on physical devices and online emulator benchmarks show that our method discovers more unique screens, covers more benchmark tasks, and more effectively rejects erroneous priors while leveraging correct ones, with accuracy gains that transfer across datasets.

cs.LG↗
arXiv2026

Bayes-Sufficient Compression Is Not Enough: How Does Communication Help Multi-Agent Systems?

Multi-agent LLM systems pair a sender with broad context and an executor with a limited local view. We study when a short message improves the executor's next decision, when raw context is preferable, and when a stronger sender helps. Our framework, \emph{receiver-relative bounded coordination}, expresses message utility as receiver gain minus protocol tax. Compression beats raw context when tax savings exceed losses from omitted information and decoder mismatch. Even \emph{Bayes-sufficient} compression can fail when a bounded executor cannot use its surface form. A three-stage decomposition separates externalization, absorption, and \emph{action closure}, explaining how errors remain after the correct content reaches the receiver. Under a single-crossing condition, sender upgrades help above a receiver-burden threshold. Across six benchmarks, the same Qwen protocol raises ContextBench joint accuracy from $0.633$ to $0.775$ but lowers ToolSandbox from $0.889$ to $0.653$. Fixed-message replay reveals closure failures despite correct artifact recovery. These results guide an inference-time selector that improves the accuracy-cost frontier on the evaluated communication regimes.

cs.LG↗
arXiv2026

Towards Understanding On-Policy Distillation through the Lens of Test-Time Scaling

On-policy distillation (OPD) has emerged as a promising post-training technique for enhancing LLM reasoning. Under the reverse KL objective, the idealized optimum of OPD aligns the student distribution with that of the teacher. When the teacher consistently outperforms the student, this naturally suggests that OPD should yield broad improvements over the pre-OPD student. However, do such improvements extend across the entire range of test-time sampling budgets? In this work, we revisit this expectation through the lens of test-time scaling by varying the sampling budget $K$ and evaluating performance with pass@$K$. Across multiple settings, we observe two distinct patterns: OPD can improve pass@$K$ at both small and large sampling budgets, but it can also improve small-budget performance while reducing large-budget pass@$K$. We show one condition that guarantees such a reversal and an idealized reverse KL counterexample where it occurs even when the teacher has higher accuracy on every problem. To choose between two candidate teachers at a target sampling budget, we propose the \textit{Teacher Advantage Score at $K$} (TAS@$K$), which can be computed before OPD training to predict which teacher will lead to a larger improvement in pass@$K$. Across three domains and thirteen benchmarks, the ordering predicted by TAS@$K$ agrees with the observed pass@$K$ improvements of the resulting OPD models in 83.6\% of experiments, providing a useful signal for teacher selection at the target pass@$K$.

cs.LG↗
arXiv2026

DE-Venus: A Data-Efficient RLVR Framework for Large Language Models

Reinforcement learning with verifiable rewards (RLVR) improves large language model reasoning, but its practical scaling is constrained by expensive on-policy rollouts and the cost of obtaining reliable targets at scale. Existing methods address sample selection, incomplete supervision, or noisy labels separately, often entangling supervision logic with distributed training and hindering controlled comparison and reuse. We present DE-Venus, a unified framework for data-efficient RLVR that treats supervision as evolving state across data preparation and policy optimization. It organizes this lifecycle into three modules: Active Data Selection allocates training and annotation budgets; Weak Supervision Construction derives learning signals from unlabeled examples; and Training-Time Supervision Refinement filters or corrects unreliable supervision. DE-Venus supports seven representative methods and a data-selection pipeline by expressing method-specific decisions as offline dataset transitions or online transformations of targets, rewards, batches, and advantages while preserving verl's distributed execution contracts. Across public benchmarks and three business scenarios, separate configurations preserve or improve model quality with only 10% of labels or as little as 13% of relevant data; selected business configurations also reduce observed convergence steps by 63%--75%. DE-Venus thus reduces annotation and training costs without sacrificing scalable RL execution.

cs.LG↗
arXiv2026

Scaling Large Reasoning Models beyond Human Supervision: A Path toward Superintelligence

Recent advances in large reasoning models (LRMs) have shown that reinforcement learning with verifiable rewards (RLVR) can substantially improve reasoning in mathematics and code, where outcomes can be checked automatically. Extending this progress to open-ended and agentic tasks remains difficult because reliable rewards are harder to obtain and direct human supervision cannot keep pace with the scale and complexity of model-generated experience. This paper studies how LRMs can continue to improve as human supervision gradually recedes from the learning loop. We examine two connected dimensions of this problem. The reward axis traces the development from per-instance human judgments to reusable verifiers and rewards that operate even without human feedback. The experience axis examines how learning can progress from human-curated tasks and environments toward self-generated curricula, constructed environments, and autonomous co-evolution. We connect these dimensions through a five-level ladder from L0 to L4 that identifies which parts of the learning process remain under continued human control. Our analysis further highlights the risks introduced by increasingly autonomous rewards and experience generation, including reward hacking, feedback drift, curriculum collapse, and environment errors. Consequently, we also provide the evaluation around three complementary objects: policy capability, feedback fidelity, and experience quality. This analysis provides a structured account of current approaches to scaling LRMs beyond human supervision and the open problems involved in developing self-sustaining learning systems toward superintelligence. Furthermore, we maintain a continuously updated \href{https://github.com/visitworld123/Awesome-Scaling-LRM-Beyond-Human-Supervision}{GitHub repository} to track the latest advances.

cs.AI↗
arXiv2026

PhysMLLMs: Spatial Priors for Unified Referring Segmentation and Grounded Reasoning of Images and Videos

Video multimodal large language models support language guided video segmentation, but they often show spatio temporal inconsistencies, e.g., jitter, drift, and identity switches. These failures are more common when targets are partly hidden or when similar objects appear nearby.One likely reason is that current training lacks explicit spatial priors, which makes it difficult to maintain stable spatial identity and shape over time. We present PhysMLLMs, a training-stage prior injection architecture that injects physics-inspired spatial continuity priors into Video MLLMs. PhysMLLMs is designed to encourage more stable object-centered representations by aligning the student global visual representation with a frozen teacher model during training. Our core mechanism, Global Representation Prior Alignment (REPA-Global), distills global visual representations from a frozen DINOv2 teacher using an offline embedding cache and a scheduled distillation plan. This design keeps inference unchanged and does not add inference time cost. Across multiple video benchmarks, PhysMLLMs improves video segmentation mask quality and cross-frame consistency, with larger gains on challenging cases involving small targets, fast motion, occlusion, distractors, and reasoning queries. On single-frame referring image segmentation and representative general VLM benchmarks, PhysMLLMs maintains comparable performance, demonstrating that the injected spatial prior improves video consistency without compromising image-level grounding or general multimodal capability. These results suggest that physics-inspired spatial prior injection can improve temporal stability while preserving general capability. The code is available at https://github.com/tusu-code/20260121-icml2026-2.git.

cs.AI↗
arXiv2026

PRISM: Streaming Human Motion Generation with Per-Joint Latent Decomposition

Text-to-motion generation has advanced with larger corpora and stronger generators, yet many models still rely on holistic frame- or clip-level latents that entangle trajectory, orientation, and articulation. This entanglement obscures body topology and forces the generator to recover kinematic structure implicitly. We present \name, a SMPL motion generation framework that factorizes motion into continuous kinematic-unit latents. A causal Motion VAE maps motion to a time-by-kinematic-unit latent manifold, and a Kinematic-Unit Flow Transformer performs text-conditioned flow matching in this structured space. Because each latent coordinate remains tied to a physical body unit, \name can use kinematic-tree rotary position encoding and kinematic-adaptive flow scheduling. We further train the generator with per-token timesteps over clean-context/noisy-target masks, enabling frame-conditioned continuation and autoregressive segment chaining within one model. Experiments first validate the representation: the kinematic-unit VAE achieves lower geometry, rotation, and feature errors than existing motion tokenizers, showing that the latent space preserves articulated structure rather than merely compressing frames. With a 1.4B-parameter generator trained only on publicly available academic motion--text data, \name outperforms all evaluated academic-data text-to-motion baselines and remains competitive with systems trained on much larger non-public motion corpora. Without task-specific retraining, the same formulation also improves prefix-conditioned generation, BABEL sequential rollout, and narrative motion composition. These results indicate that kinematic-unit latent factorization provides an effective generation substrate for controllable SMPL motion synthesis. Code will be released at https://github.com/ZeyuLing/PRISM.

cs.CV↗
arXiv2026

Model Inversion Attacks: A Survey of Approaches and Countermeasures

Deep neural networks have enabled numerous studies and applications on both Euclidean data, such as images and text, and non-Euclidean data, such as graphs. Because these networks may process private data, their deployment raises concerns about privacy leakage. Model inversion attacks (MIAs) exploit access to a trained model to reconstruct training examples or infer privacy-sensitive characteristics represented by the model. The effectiveness of MIAs has been demonstrated in various domains, including images, text, and graphs. These attacks highlight the vulnerability of neural networks and raise awareness about the risk of privacy leakage within the research community. This survey provides a threat-model- and assumption-aware synthesis of attacks and defenses. We compare exposed interfaces, attacker knowledge, reconstruction priors and targets, failure modes, deployment constraints, and privacy-utility trade-offs, while highlighting modeling principles, optimization challenges, and future directions. We also maintain an evolving repository of relevant research at https://github.com/AndrewZhou924/Awesome-model-inversion-attack.

cs.LG↗
arXiv2026

Adversarial Purification by Consistency-aware Latent Space Optimization on Data Manifolds

Deep neural networks (DNNs) are vulnerable to adversarial samples crafted by adding imperceptible perturbations to clean data, potentially leading to incorrect and dangerous predictions. Adversarial purification has been an effective means to improve DNNs robustness by removing these perturbations before feeding the data into the model. However, it faces significant challenges in preserving key structural and semantic information of data, as the imperceptible nature of adversarial perturbations makes it hard to avoid over-correcting, which can destroy important information and degrade model performance. In this paper, we break away from traditional adversarial purification methods by focusing on the clean data manifold. To this end, we reveal that samples generated by a well-trained generative model are close to clean ones but far from adversarial ones. Leveraging this insight, we propose Consistency Model-based Adversarial Purification (CMAP), which optimizes vectors within the latent space of a pre-trained consistency model to generate samples for restoring clean data. Specifically, 1) we propose a Perceptual consistency restoration mechanism by minimizing the discrepancy between generated samples and input samples in both pixel and perceptual spaces. 2) To maintain the optimized latent vectors within the valid data manifold, we introduce a Latent distribution consistency constraint strategy to align generated samples with the clean data distribution. 3) We also apply a Latent vector consistency prediction scheme via an ensemble approach to enhance prediction reliability. CMAP fundamentally addresses adversarial perturbations at their source, providing a robust purification. Extensive experiments on CIFAR-10 and ImageNet-100 show that our CMAP significantly enhances robustness against strong adversarial attacks while preserving high natural accuracy.

cs.LG↗
arXiv2026

Charge Tunable Optical Nonlinearity of Moiré Exciton-Polaritons

Transition metal dichalcogenides represent a versatile platform to study strong light-matter interactions based on excitons and electrons in ordered lattices. Twist-engineering of moiré structures further enables the manipulation of the polaritonic nonlinearities via engineering the exciton landscape on the nanoscale. In this work, we demonstrate in-situ control of the optical saturation-based nonlinearity of moiré exciton-polaritons by phase space restriction via charge doping. Strong exciton-photon coupling is established in a gate-controllable MoTe$_2$-MoSe$_2$ heterobilayer, embedded in a spectrally-tunable open cavity. A small gate voltage can effectively lower the necessary polariton density by one order of magnitude to achieve a similar nonlinear saturation effect as in the charge-neutral case. Our microscopic description successfully explains the observed phenomena in the framework of Pauli blocking for the moiré superlattices with charge preoccupation.

cond-mat.mes-hall↗
arXiv2026

Generated Images Are Easier to Forget: A Machine Unlearning Perspective for Synthetic Image Detection

Robust detection of generated images is critical to counter the misuse of generative models. Existing methods primarily depend on learning from human-annotated training datasets, limiting their generalization to unseen distributions. In contrast, large-scale vision models (LVMs) pre-trained on web-scale datasets exhibit exceptional generalization power through exposure to diverse distributions, offering a transformative paradigm for this task. However, our experimental results reveal that LVMs pre-trained on natural-image-dominated data can effectively capture the features of both natural and generated images, yielding comparably low losses and thus limited discriminative capacity between them. This prompts a key question: When and how do LVMs exhibit different behaviors when capturing features of natural and generated images? This investigation reveals an insight: during unlearning, LVMs exhibit disparate forgetting dynamics with feature degradation for generated images escalating faster than natural ones. Inspired by the disparate dynamics, we introduce two detection methods: 1) data-free detection, which prunes model parameters to induce unlearning without data access, and 2) data-driven detection, which optimizes LVMs to unlearn knowledge tied to generated images. Extensive experiments conducted on various benchmarks demonstrate that our unlearning-based approach outperforms conventional detection methods. By recasting the detection task as a problem of machine unlearning, our work establishes a new paradigm for generated image detection.

cs.CV↗
arXiv2026

Concept Concentration for Faithful Representation Intervention

Representation intervention aims to localize and modify the representations that encode the underlying concepts in large language models (LLMs) to elicit the aligned and expected behaviors. Despite the empirical success, it has never been examined whether one could localize the faithful concepts for intervention. In this work, we explore the question in safety alignment. If the interventions are faithful, the intervened LLMs should erase the harmful concepts and be robust to both in-distribution adversarial prompts and the out-of-distribution (OOD) jailbreaks. While it is feasible to erase harmful concepts without degrading the benign utility of LLMs in linear settings, we show that it is infeasible in the general non-linear setting. To tackle the issue, we propose COncept ConcentrAtion (COCA). COCA refactors the training data with an explicit reasoning process, which first identifies the potential unsafe concepts and then decides the responses. Essentially, COCA simplifies the decision boundary between harmful and benign representations, enabling more effective linear erasure. Extensive experiments with multiple representation intervention methods and model architectures demonstrate that COCA significantly reduces both in-distribution and OOD jailbreak success rates, and meanwhile maintaining strong performance on regular tasks such as math and code generation. Our code is publicly available at: https://github.com/tmlr-group/COCA .

cs.LG↗
arXiv2026

Direct Measurement of Exciton Dispersion in the Long-Wavelength Limit

Exciton dispersion, which governs the propagation, scattering and radiative decay of electron-hole pairs, is essential to optoelectronics and quantum materials. In two-dimensional systems, weakened dielectric screening and long-range electron-hole exchange are predicted to induce nonanalytic exciton dispersion in the long-wavelength limit. However, direct quantitative characterization of its dimensional evolution remains lacking, especially in the ultralow-q regime (q < 0.02 $Å^{-1}$). Here we employ defocus-engineered momentum-resolved electron energy-loss spectroscopy in scanning transmission electron microscopy, achieving an ultrahigh momentum resolution of 0.0002 $Å^{-1}$. Using freestanding hBN as a prototypical platform, we resolve layer-dependent exciton dispersion and quantify its characteristic crossover momentum and group velocity in the long-wavelength limit. With increasing thickness, the nonanalytic linear-dispersion regime is progressively compressed, manifested by a reduction in characteristic crossover momentum q_c from $1.82 \times 10^{-1} Å^{-1}$ in the monolayer to $3.0 \times 10^{-1} Å^{-1}$ in 25 layers. Meanwhile, the low-q group velocity increases from $2.0 \times 10^{-3} c$ to $2.9 \times 10^{-2} c$, before the dispersion ultimately approaches the bulk-like parabolic limit. We further examine how the exciton band structure of monolayer hBN responds to its surrounding environment, including temperature, adjacent graphene layers, and interlayer twist in BN/graphene heterostructures. These findings uncover the fundamental physics of low-dimensional excitons, deliver valuable guidance for modulating exciton transport, diffusion and quasiparticle coupling in layered quantum materials, and establish a powerful experimental route to explore low-dimensional exciton physics.

cond-mat.mes-hall↗
arXiv2026

Memory-Augmented Multimodal Large Language Models for Small Object Understanding in Streaming Aerial Videos

Language-guided aerial perception aims to understand user-specified tiny targets in complex unmanned aerial vehicle (UAV) scenes. In real UAV deployment, the UAV must respond while it flies, so such perception runs in an online streaming manner, where frames arrive sequentially and the model responds to each one without access to future frames. However, applying current Multimodal Large Language Models (MLLMs) to this setting raises two challenges. First, targets viewed from the air are often tiny, yet the visual compression in existing MLLMs treats all regions equally and discards their fine-grained details. Second, understanding a continuous stream requires past-frame context, yet retaining the entire history is infeasible on resource-constrained onboard hardware, whereas discarding it causes the target to drift or disappear. We address the tiny object and streaming challenges from both data and method perspectives. From the data perspective, we present \textbf{DroneEyes}, the \textbf{first} pixel-level and open-vocabulary referring-segmentation dataset for tiny aerial targets, comprising $2,140$ high-definition videos and $176,623$ pairs across Object Description and Referring Expression tasks, with dense per-frame masks. From the method perspective, we propose \textbf{SkyAnchor}, an MLLM with two designs to the above challenges: a Semantics-Aware Token Router that preserves small-target under a reduced visual-token budget, and a Hierarchical Memory Bank that keeps the target consistently understood on streams.

cs.CV↗
arXiv2026

Machine Learning Potential-Driven Molecular Dynamics Simulations of Dehydrogenation in Pristine and Doped MgH$_2$

Machine learning potential-driven molecular dynamics simulations (ML-MD) were employed to provide atomistic insights into the dehydrogenation kinetics of pristine and doped MgH2. Through systematic investigation of distinct surface orientations, the MgH2 (100) surface was identified as the most active low-index surface for hydrogen release. For pristine MgH2, our simulations revealed a novel H2 formation mechanism characterized by H2 generation in the subsurface region followed by diffusion to the surface for desorption, highlighting the critical role of subsurface processes beyond conventional surface-driven pathways. Comprehensive screening of 22 doping elements identified Ni as the most effective dopant. Among several descriptors, machine learning analysis identified the time-coupled Miedema electron density as the critical descriptor, underscoring the role of electronic properties. Consequently, a volcano-shaped relationship was uncovered between the intrinsic Miedema electron density ( nws ) and total hydrogen release (optimal window: 4.0 < nws < 5.4x10-2 e/bohr3). Dopants within this range serve a dual function: acting as thermodynamic sinks for H attraction while maintaining a balanced interaction strength to facilitate H-H coupling and H2 release. This atomistic-level validation provides strong theoretical support for the experimentally observed "hydrogen pump" effect of catalytic phases. The present study demonstrates the strong capability of ML- MD in navigating through complex catalytic mechanisms and establishing quantitative property-activity relationships, providing a robust framework for rational design of high-performance catalysts for MgH2 and other hydrogen storage materials.

cond-mat.mtrl-sci↗