arXiv subjects

Bohan Zhang

Publications and source records attributed to Bohan Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

OpenAI o1 System Card

The o1 model series is trained with large-scale reinforcement learning to reason using chain of thought. These advanced reasoning capabilities provide new avenues for improving the safety and robustness of our models. In particular, our models can reason about our safety policies in context when responding to potentially unsafe prompts, through deliberative alignment. This leads to state-of-the-art performance on certain benchmarks for risks such as generating illicit advice, choosing stereotyped responses, and succumbing to known jailbreaks. Training models to incorporate a chain of thought before answering has the potential to unlock substantial benefits, while also increasing potential risks that stem from heightened intelligence. Our results underscore the need for building robust alignment methods, extensively stress-testing their efficacy, and maintaining meticulous risk management protocols. This report outlines the safety work carried out for the OpenAI o1 and OpenAI o1-mini models, including safety evaluations, external red teaming, and Preparedness Framework evaluations.

cs.AI↗

arXiv2026

PACO: Proxy-Task Alignment and Online Calibration for On-the-Fly Category Discovery

On-the-Fly Category Discovery (OCD) requires a model, trained on an offline support set, to recognize known classes while discovering new ones from an online streaming sequence. Existing methods focus heavily on offline training. They aim to learn discriminative representations on the support set so that novel classes can be separated at test time. However, their discovery mechanism at inference is typically reduced to a single threshold. We argue that this paradigm is fundamentally flawed as OCD is not a static classification problem, but a dynamic process. The model must continuously decide 1) whether a sample belongs to a known class, 2) matches an existing novel category, or 3) should initiate a new one. Moreover, prior methods treat the support set as fixed knowledge. They do not update their decision boundaries as new evidence arrives during inference. This leads to unstable and inconsistent category formation. Our experiments confirm these issues. With properly calibrated and adaptive thresholds, substantial improvements can be achieved, even without changing the representation. Motivated by this, we propose PACO, a support-set-calibrated, tree-structured online decision framework. The framework models inference as a sequence of hierarchical decisions, including known-class routing, birth-aware novel assignment, and attach-versus-create operations over a dynamic prototype memory. Furthermore, we simulate the proxy discovery process to initialize the thresholds during offline training to align with inference. Thresholds are continuously updated during inference using mature novel prototypes. Importantly, PACO requires no heavy training and no dataset-specific tuning. It can be directly integrated into existing OCD pipelines as an inference-time module. Extensive experiments show significant improvements over SOTA baselines across seven benchmarks.

cs.CV↗

arXiv2026

AutonomyLens: A Self-Evolving Simulation-Based Testing Loop for Autonomous Systems

Software engineering practices for validating autonomous cyber-physical systems (e.g., Uncrewed Aerial Vehicles) remain fragmented across scenario design, simulation execution, and telemetry analysis, limiting traceability between requirements, tests, and evidence. This fragmentation reduces reproducibility, slows debugging and iteration, and hinders systematic assurance under complex and evolving environmental conditions. We present AutonomyLens, an LLM-driven framework that integrates scenario specification, simulation execution, and telemetry analysis into a unified validation workflow. AutonomyLens enables developers to translate high-level validation intent into executable, temporally evolving scenarios, automatically run simulations, and perform context-aware analysis of resulting system behavior. The framework introduces (i) a structured representation for mission-level scenarios, (ii) an automated execution pipeline, (iii) analysis mechanisms that align telemetry with scenario context to produce actionable insights, and (iv) counterfactual scenario generation that closes the loop by refining and synthesizing new test cases from observed failures. We describe the early-stage design of AutonomyLens, discuss key challenges in building integrated validation workflows for autonomous systems, and outline how such an approach can improve traceability, reproducibility, and scalability in autonomy validation.

cs.SE↗

arXiv2026

Sparse-RL: Breaking the Memory Wall in LLM Reinforcement Learning via Stable Sparse Rollouts

Reinforcement Learning (RL) has become essential for eliciting complex reasoning capabilities in Large Language Models (LLMs). However, the substantial memory overhead of storing Key-Value (KV) caches during long-horizon rollouts acts as a critical bottleneck, often prohibiting efficient training on limited hardware. While existing KV compression techniques offer a remedy for inference, directly applying them to RL training induces a severe policy mismatch, leading to catastrophic performance collapse. To address this, we introduce Sparse-RL empowers stable RL training under sparse rollouts. We show that instability arises from a fundamental policy mismatch among the dense old policy, the sparse sampler policy, and the learner policy. To mitigate this issue, Sparse-RL incorporates Sparsity-Aware Rejection Sampling and Importance-based Reweighting to correct the off-policy bias introduced by compression-induced information loss. Experimental results show that Sparse-RL reduces rollout overhead compared to dense baselines while preserving the performance. Furthermore, Sparse-RL inherently implements sparsity-aware training, significantly enhancing model robustness during sparse inference deployment. The corresponding training data and code are publicly available on the repository.

cs.LG↗

arXiv2026

Learning through Creation: A Hash-Free Framework for On-the-Fly Category Discovery

On-the-Fly Category Discovery (OCD) aims to recognize known classes while simultaneously discovering emerging novel categories during inference, using supervision only from known classes during offline training. Existing approaches rely either on fixed label supervision or on diffusion-based augmentations to enhance the backbone, yet none of them explicitly train the model to perform the discovery task required at test time. It is fundamentally unreasonable to expect a model optimized on limited labeled data to carry out a qualitatively different discovery objective during inference. This mismatch creates a clear optimization misalignment between the offline learning stage and the online discovery stage. In addition, prior methods often depend on hash-based encodings or severe feature compression, which further limits representational capacity. To address these issues, we propose Learning through Creation (LTC), a fully feature-based and hash-free framework that injects novel-category awareness directly into offline learning. At its core is a lightweight, online pseudo-unknown generator driven by kernel-energy minimization and entropy maximization (MKEE). Unlike previous methods that generate synthetic samples once before training, our generator evolves jointly with the model dynamics and synthesizes pseudo-novel instances on the fly at negligible cost. These samples are incorporated through a dual max-margin objective with adaptive thresholding, strengthening the model's ability to delineate and detect unknown regions through explicit creation. Extensive experiments across seven benchmarks show that LTC consistently outperforms prior work, achieving improvements ranging from 1.5 percent to 13.1 percent in all-class accuracy. The code is available at https://github.com/brandinzhang/LTC

cs.CV↗

arXiv2026

From Folding Mechanics to Robotic Function: A Unified Modeling Framework for Compliant Origami

Origami inspired architectures offer a powerful route toward lightweight, reconfigurable, and programmable robotic systems. Yet, a unified mechanics framework capable of seamlessly bridging rigid folding, elastic deformation, and stability driven transitions in compliant origami remains lacking. Here, we introduce a geometry consistent modeling framework based on discrete differential geometry (DDG) that unifies panel elasticity and crease rotation within a single variational formulation. By embedding crease panel coupling directly into a mid edge geometric discretization, the framework naturally captures rigid folding limits, distributed bending, multistability, and nonlinear dynamic snap through within one mechanically consistent structure. This unified description enables programmable control of stability and deformation across rigid and compliant regimes, allowing origami structures to transition from static folding mechanisms to active robotic modules. An implicit dynamic formulation incorporating gravity, contact, friction, and magnetic actuation further supports strongly coupled multiphysics simulations. Through representative examples spanning single fold bifurcation, deployable Miura membranes, bistable Waterbomb modules, and Kresling based crawling robots, we demonstrate how geometry driven mechanics directly informs robotic functionality. This work establishes discrete differential geometry as a foundational design language for intelligent origami robotics, enabling predictive modeling, stability programming, and mechanics guided robotic actuation within a unified computational platform.

cs.RO↗

arXiv2026

CoSA: Compressed Sensing-Based Adaptation of Large Language Models

Parameter-Efficient Fine-Tuning (PEFT) has emerged as a practical paradigm for adapting large language models (LLMs) without updating all parameters. Most existing approaches, such as LoRA and PiSSA, rely on low-rank decompositions of weight updates. However, the low-rank assumption may restrict expressivity, particularly in task-specific adaptation scenarios where singular values are distributed relatively uniformly. To address this limitation, we propose CoSA (Compressed Sensing-Based Adaptation), a new PEFT method extended from compressed sensing theory. Instead of constraining weight updates to a low-rank subspace, CoSA expresses them through fixed random projection matrices and a compact learnable core. We provide a formal theoretical analysis of CoSA as a synthesis process, proving that weight updates can be compactly encoded into a low-dimensional space and mapped back through random projections. Extensive experimental results show that CoSA provides a principled perspective for efficient and expressive multi-scale model adaptation. Specifically, we evaluate CoSA on 10 diverse tasks, including natural language understanding and generation, employing 5 models of different scales from RoBERTa, Llama, and Qwen families. Across these settings, CoSA consistently matches or outperforms state-of-the-art PEFT methods.

cs.LG↗

arXiv2026

ContextBench: A Benchmark for Context Retrieval in Coding Agents

LLM-based coding agents have shown strong performance on automated issue resolution benchmarks, yet existing evaluations largely focus on final task success, providing limited insight into how agents retrieve and use code context during problem solving. We introduce ContextBench, a process-oriented evaluation of context retrieval in coding agents. ContextBench consists of 1,136 issue-resolution tasks from 66 repositories across eight programming languages, each augmented with human-annotated gold contexts. We further implement an automated evaluation framework that tracks agent trajectories and measures context recall, precision, and efficiency throughout issue resolution. Using ContextBench, we evaluate four frontier LLMs and five coding agents. Our results show that sophisticated agent scaffolding yields only marginal gains in context retrieval ("The Bitter Lesson" of coding agents), LLMs consistently favor recall over precision, and substantial gaps exist between explored and utilized context. ContextBench augments existing end-to-end benchmarks with intermediate gold-context metrics that unbox the issue-resolution process. These contexts offer valuable intermediate signals for guiding LLM reasoning in software tasks.

cs.LG↗

arXiv2026

Neural Networks Learn Generic Multi-Index Models Near Information-Theoretic Limit

In deep learning, a central issue is to understand how neural networks efficiently learn high-dimensional features. To this end, we explore the gradient descent learning of a general Gaussian Multi-index model $f(\boldsymbol{x})=g(\boldsymbol{U}\boldsymbol{x})$ with hidden subspace $\boldsymbol{U}\in \mathbb{R}^{r\times d}$, which is the canonical setup to study representation learning. We prove that under generic non-degenerate assumptions on the link function, a standard two-layer neural network trained via layer-wise gradient descent can agnostically learn the target with $o_d(1)$ test error using $\widetilde{\mathcal{O}}(d)$ samples and $\widetilde{\mathcal{O}}(d^2)$ time. The sample and time complexity both align with the information-theoretic limit up to leading order and are therefore optimal. During the first stage of gradient descent learning, the proof proceeds via showing that the inner weights can perform a power-iteration process. This process implicitly mimics a spectral start for the whole span of the hidden subspace and eventually eliminates finite-sample noise and recovers this span. It surprisingly indicates that optimal results can only be achieved if the first layer is trained for more than $\mathcal{O}(1)$ steps. This work demonstrates the ability of neural networks to effectively learn hierarchical functions with respect to both sample and time efficiency.

stat.ML↗

arXiv2026

NurValues: Real-World Nursing Values Evaluation for Large Language Models in Clinical Context

While LLMs have demonstrated medical knowledge and conversational ability, their deployment in clinical practice raises new risks: patients may place greater trust in LLM-generated responses than in nurses' professional judgments, potentially intensifying nurse-patient conflicts. Such risks highlight the urgent need of evaluating whether LLMs align with the core nursing values upheld by human nurses. This work introduces the first benchmark for nursing value alignment, consisting of five core value dimensions distilled from international nursing codes: Altruism, Human Dignity, Integrity, Justice, and Professionalism. We define two-level tasks on the benchmark, considering the two characteristics of emerging nurse-patient conflicts. The Easy-Level dataset consists of 2,200 value-aligned and value-violating instances, which are collected through a five-month longitudinal field study across three hospitals of varying tiers; The Hard-Level dataset is comprised of 2,200 dialogue-based instances that embed contextual cues and subtle misleading signals, which increase adversarial complexity and better reflect the subjectivity and bias of narrators in the context of emerging nurse-patient conflicts. We evaluate a total of 23 SoTA LLMs on their ability to align with nursing values, and find that general LLMs outperform medical ones, and Justice is the hardest value dimension. As the first real-world benchmark for healthcare value alignment, NurValues provides novel insights into how LLMs navigate ethical challenges in clinician-patient interactions.

cs.CL↗

arXiv2026

Who Owns the Text? Design Patterns for Preserving Authorship in AI-Assisted Writing

AI writing assistants can reduce effort and improve fluency, but they may also weaken writers' sense of authorship. We study this tension with an ownership-aware co-writing editor that offers on-demand, sentence-level suggestions and tests two common design choices: persona-based coaching and style personalization. In an online study (N=176), participants completed three professional writing tasks: an email without AI help, a proposal with generic AI suggestions, and a cover letter with persona-based coaching, while half received suggestions tailored to a brief sample of their prior writing. Across the two AI-assisted tasks, psychological ownership dropped relative to unassisted writing (about 0.85-1.0 points on a 7-point scale), even as cognitive load decreased (about 0.9 points) and quality ratings stayed broadly similar overall. Persona coaching did not prevent the ownership decline. Style personalization partially restored ownership (about +0.43) and increased AI incorporation in text (+5 percentage points). We distill five design patterns: on-demand initiation, micro-suggestions, voice anchoring, audience scaffolds, and point-of-decision provenance, to guide authorship-preserving writing tools.

cs.HC↗

arXiv2025

Thermal conductivities of monolayer graphene oxide from machine learning molecular dynamics simulations

Graphene oxide (GO) exhibits rich chemical heterogeneity that strongly influences its structural, thermal, and mechanical properties, yet quantitatively linking reduction chemistry to heat transport remains challenging. In this work, we develop a machine-learned neuroevolution potential (NEP) trained on an existing density functional theory dataset (\textit{Angew.\ Chem.\ Int.\ Ed.}, \textbf{63} , e202410088 (2024)), achieving reasonable accuracy at a computational cost much lower than the existing machine-learned and empirical potentials. Leveraging this potential, we perform large-scale molecular dynamics (MD) simulations to model the thermal reduction of GO across realistic structural domains. Using the homogeneous nonequilibrium MD method with a proper quantum-statistical correction scheme, we find that reduced GO exhibits strongly suppressed thermal conductivities, ranging from a few to tens of Wm$^{-1}$K$^{-1}$, substantially lower than pristine GO without defects and far below graphene. Moreover, the thermal conductivity of reduced GO increases moderately with increasing OH/O ratio, except at the highest oxidation level (O/C=0.5) where this trend inverts, while decreasing significantly with increasing O/C ratio, a trend strongly correlated with the fraction of recovered graphene-like structures. Our work provides a computationally tractable and predictive atomistic machine learning framework for exploring how chemical structure governs heat transport in heterogeneous carbon materials.

physics.chem-ph↗

arXiv2025

Brain-Muscle Atlas: A novel framework for Motor Brain-Computer Interfaces

Motor brain-computer interfaces (BCIs) enable the control of external devices by decoding neural signals. However, most existing systems rely on a direct "brain-machine" mapping, overlooking the hierarchical physiological pathway of natural movement, namely the "brain-muscle-joint" cascade. Due to the lack of explicit modeling and enhancement of this pathway, current systems are often constrained by the low amplitude and high noise of EEG signals, resulting in motor outputs that are unstable, discontinuous, and insufficiently natural.To address these limitations, this study introduces the concept of a brain-muscle atlas, designed to systematically characterize the mapping between motor cortical activity and corresponding muscle activation, thereby establishing a movement decoding framework that better aligns with neuromuscular physiology. Using synchronously recorded EEG-EMG data, we constructed the first brain-muscle atlas for elbow flexion-extension, achieving a structured mapping from cortical activity to muscle activation.Offline experiments demonstrate that the proposed atlas accurately reconstructs the temporal activation patterns of primary elbow agonists, achieving a maximum correlation coefficient of 0.8314, thereby validating its ability to capture cortical-muscular mapping. Furthermore, by leveraging atlas-derived muscle activation representations, we enabled continuous real-time control of a virtual elbow joint. All ten participants successfully completed the online flexion-extension task, indicating that the system robustly extracts motor intent even under low-SNR EEG conditions.

cs.HC↗

arXiv2025

List Replicable Reinforcement Learning

Replicability is a fundamental challenge in reinforcement learning (RL), as RL algorithms are empirically observed to be unstable and sensitive to variations in training conditions. To formally address this issue, we study \emph{list replicability} in the Probably Approximately Correct (PAC) RL framework, where an algorithm must return a near-optimal policy that lies in a \emph{small list} of policies across different runs, with high probability. The size of this list defines the \emph{list complexity}. We introduce both weak and strong forms of list replicability: the weak form ensures that the final learned policy belongs to a small list, while the strong form further requires that the entire sequence of executed policies remains constrained. These objectives are challenging, as existing RL algorithms exhibit exponential list complexity due to their instability. Our main theoretical contribution is a provably efficient tabular RL algorithm that guarantees list replicability by ensuring the list complexity remains polynomial in the number of states, actions, and the horizon length. We further extend our techniques to achieve strong list replicability, bounding the number of possible policy execution traces polynomially with high probability. Our theoretical result is made possible by key innovations including (i) a novel planning strategy that selects actions based on lexicographic order among near-optimal choices within a randomly chosen tolerance threshold, and (ii) a mechanism for testing state reachability in stochastic environments while preserving replicability. Finally, we demonstrate that our theoretical investigation sheds light on resolving the \emph{instability} issue of RL algorithms used in practice. In particular, we show that empirically, our new planning strategy can be incorporated into practical RL frameworks to enhance their stability.

cs.LG↗

arXiv2025

System Filter-Based Common Components Modeling for Cross-Subject EEG Decoding

Brain-computer interface (BCI) technology enables direct communication between the brain and external devices through electroencephalography (EEG) signals. However, existing decoding models often mix common and personalized components, leading to interference from individual variability that limits cross-subject decoding performance. To address this issue, this paper proposes a system filter that extends the concept of signal filtering to the system level. The method expands a system into its spectral representation, selectively removes unnecessary components, and reconstructs the system from the retained target components, thereby achieving explicit system-level decomposition and filtering. We further integrate the system filter into a Cross-Subject Decoding framework based on the System Filter (CSD-SF) and evaluate it on the four-class motor imagery (MI) task of the BCIC IV 2a dataset. Personalized models are transformed into relation spectrums, and statistical testing across subjects is used to remove personalized components. The remaining stable relations, representing common components across subjects, are then used to construct a common model for cross-subject decoding. Experimental results show an average improvement of 3.28% in decoding accuracy over baseline methods, demonstrating that the proposed system filter effectively isolates stable common components and enhances model robustness and generalizability in cross-subject EEG decoding.

q-bio.NC↗

arXiv2025

ArchMap: Arch-Flattening and Knowledge-Guided Vision Language Model for Tooth Counting and Structured Dental Understanding

A structured understanding of intraoral 3D scans is essential for digital orthodontics. However, existing deep-learning approaches rely heavily on modality-specific training, large annotated datasets, and controlled scanning conditions, which limit generalization across devices and hinder deployment in real clinical workflows. Moreover, raw intraoral meshes exhibit substantial variation in arch pose, incomplete geometry caused by occlusion or tooth contact, and a lack of texture cues, making unified semantic interpretation highly challenging. To address these limitations, we propose ArchMap, a training-free and knowledge-guided framework for robust structured dental understanding. ArchMap first introduces a geometry-aware arch-flattening module that standardizes raw 3D meshes into spatially aligned, continuity-preserving multi-view projections. We then construct a Dental Knowledge Base (DKB) encoding hierarchical tooth ontology, dentition-stage policies, and clinical semantics to constrain the symbolic reasoning space. We validate ArchMap on 1060 pre-/post-orthodontic cases, demonstrating robust performance in tooth counting, anatomical partitioning, dentition-stage classification, and the identification of clinical conditions such as crowding, missing teeth, prosthetics, and caries. Compared with supervised pipelines and prompted VLM baselines, ArchMap achieves higher accuracy, reduced semantic drift, and superior stability under sparse or artifact-prone conditions. As a fully training-free system, ArchMap demonstrates that combining geometric normalization with ontology-guided multimodal reasoning offers a practical and scalable solution for the structured analysis of 3D intraoral scans in modern digital orthodontics.

cs.CV↗

arXiv2025

Representational power of selected neural network quantum states in second quantization

Neural network quantum states emerge as a promising tool for solving quantum many-body problems. However, its successes and limitations are still not well-understood in particular for Fermions with complex sign structures. Based on our recent work [J. Chem. Theory Comput. 21, 10252-10262 (2025)], we generalizes the restricted Boltzmann machine to a more general class of states for Fermions, formed by product of `neurons' and hence will be referred to as neuron product states (NPS). NPS builds correlation in a very different way, compared with the closely related correlator product states (CPS) [H. J. Changlani, et al. Phys. Rev. B, 80, 245116 (2009)], which use full-rank local correlators. In constrast, each correlator in NPS contains long-range correlations across all the sites, with its representational power constrained by the simple function form. We prove that products of such simple nonlocal correlators can approximate any wavefunction arbitrarily well under certain mild conditions on the form of activation functions. In addition, we also provide elementary proofs for the universal approximation capabilities of feedforward neural network (FNN) and neural network backflow (NNBF) in second quantization. Together, these results provide a deeper insight into the neural network representation of many-body wavefunctions in second quantization.

quant-ph↗

arXiv2025

Nonlinear shift along the sensorimotor-association-axis in brain responses to task performance

In the literature of cognitive neuroscience, researchers tend to assume a linear relationship between brain activation level and task performance; however, controversial findings have been reported in participants at different ages and different proficiency levels. Therefore, there may be a non-linear relationship between task performance and brain activation if a full range of task performance is considered. In the current study, using the Human Connectome Project (HCP) dataset we examined the relationship between brain activation and working memory performance in two conditions (i.e. faces and places). We found a gradual change from a U-shaped relationship to an inverted U-shaped relationship along the sensorimotor-association (S-A) axis in the face condition. In other words, in low-order sensorimotor areas, it is U-shaped and in the high-order prefrontal and association areas, it is inverted U-shaped, which suggests different properties in the encoding/representation region and in the cognitive calculation regions. However, in the place condition, such a shift is missing, presumably because most of the regions that are sensitive to task performance in the place condition are in the lower end of the S-A axis. Taken together, our study revealed a novel difference of functional property in response to task performance in the sensorimotor areas versus the association areas.

q-bio.NC↗