arXiv subjects

Stephen Tu

Publications and source records attributed to Stephen Tu.

At least 19 recordsLinked to original sources

arXiv2026

Forward Trajectory Steering for Hamilton-Jacobi Reachability Analysis

Hamilton-Jacobi (HJ) reachability provides a mathematically rigorous framework for safe control of dynamical systems, but its practical application is bottlenecked by the computational complexity of solving Hamilton-Jacobi-Isaacs variational inequality PDEs in high dimensions. Physics-informed neural networks (PINNs) have recently emerged as a promising alternative to classical mesh-based solvers, yet their performance is highly sensitive to the choice of collocation sampling. In order to learn accurate safety value functions, existing PINNs-based HJ reachability solvers must rely on complex training pipelines and auxiliary supervision. In this work, we propose STEER2REACH (S2R), a PINNs-based HJ reachability solver that requires minimal modification on top of standard PINNs training. S2R's key contribution is a lightweight, low-overhead adaptive collocation sampling distribution constructed by steering forward trajectories using a combination of the optimal control and disturbance signals induced by the current value function, with injected stochastic exploration noise. We demonstrate that despite its simplicity, S2R achieves competitive--and in some cases improved--performance on safety metrics while reducing relative L2 error across a range of reachability benchmarks compared with SoTA MPC-guided HJ reachability solvers, all without requiring multi-stage training or MPC-based supervision.

eess.SY↗

arXiv2026

CLT-Optimal Parameter Error Bounds for Linear System Identification

There has been remarkable progress over the past decade in establishing finite-sample, non-asymptotic bounds on recovering unknown system parameters from observed system behavior. Surprisingly, however, we show that the current state-of-the-art bounds do not accurately capture the statistical complexity of system identification, even in the most fundamental setting of estimating a discrete-time linear dynamical system (LDS) via ordinary least-squares regression (OLS). Specifically, we utilize asymptotic normality to identify classes of problem instances for which current bounds overstate the squared parameter error, in both spectral and Frobenius norm, by a factor of the state-dimension of the system. Informed by this discrepancy, we then sharpen the OLS parameter error bounds via a novel second-order decomposition of the parameter error, where crucially the lower-order term is a matrix-valued martingale that we show correctly captures the CLT scaling. From our analysis we obtain finite-sample bounds for both (i) stable systems and (ii) the many-trajectories setting that match the instance-specific optimal rates up to constant factors in Frobenius norm, and polylogarithmic state-dimension factors in spectral norm.

stat.ML↗

arXiv2026

On the Asymptotics of Self-Supervised Pre-training: Two-Stage M-Estimation and Representation Symmetry

Self-supervised pre-training, where large corpora of unlabeled data are used to learn representations for downstream fine-tuning, has become a cornerstone of modern machine learning. While a growing body of theoretical work has begun to analyze this paradigm, existing bounds leave open the question of how sharp the current rates are, and whether they accurately capture the complex interaction between pre-training and fine-tuning. In this paper, we address this gap by developing an asymptotic theory of pre-training via two-stage M-estimation. A key challenge is that the pre-training estimator is often identifiable only up to a group symmetry, a feature common in representation learning that requires careful treatment. We address this issue using tools from Riemannian geometry to study the intrinsic parameters of the pre-training representation, which we link with the downstream predictor through a notion of orbit-invariance, precisely characterizing the limiting distribution of the downstream test risk. We apply our main result to several case studies, including spectral pre-training, factor models, and Gaussian mixture models, and obtain substantial improvements in problem-specific factors over prior art when applicable.

cs.LG↗

arXiv2026

Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisons

General-purpose robot reward models are typically trained to predict absolute task progress from expert demonstrations, providing only local, frame-level supervision. While effective for expert demonstrations, this paradigm scales poorly to large-scale robotics datasets where failed and suboptimal trajectories are abundant and assigning dense progress labels is ambiguous. We introduce Robometer, a scalable reward modeling framework that combines intra-trajectory progress supervision with inter-trajectory preference supervision. Robometer is trained with a dual objective: a frame-level progress loss that anchors reward magnitude on expert data, and a trajectory-comparison preference loss that imposes global ordering constraints across trajectories of the same task, enabling effective learning from both real and augmented failed trajectories. To support this formulation at scale, we curate RBM-1M, a reward-learning dataset comprising over one million trajectories spanning diverse robot embodiments and tasks, including substantial suboptimal and failure data. Across benchmarks and real-world evaluations, Robometer learns more generalizable reward functions than prior methods and improves robot learning performance across a diverse set of downstream applications. Code, model weights, and videos at https://robometer.github.io/.

cs.RO↗

arXiv2026

CrispEdit: Low-Curvature Projections for Scalable Non-Destructive LLM Editing

A central challenge in large language model (LLM) editing is capability preservation: methods that successfully change targeted behavior can quietly game the editing proxy and corrupt general capabilities, producing degenerate behaviors reminiscent of proxy/reward hacking. We present CrispEdit, a scalable and principled second-order editing algorithm that treats capability preservation as an explicit constraint, unifying and generalizing several existing editing approaches. CrispEdit formulates editing as constrained optimization and enforces the constraint by projecting edit updates onto the low-curvature subspace of the capability-loss landscape. At the crux of CrispEdit is expressing capability constraint via Bregman divergence, whose quadratic form yields the Gauss-Newton Hessian exactly and even when the base model is not trained to convergence. We make this second-order procedure efficient at the LLM scale using Kronecker-factored approximate curvature (K-FAC) and a novel matrix-free projector that exploits Kronecker structure to avoid constructing massive projection matrices. Across standard model-editing benchmarks, CrispEdit achieves high edit success while keeping capability degradation below 1% on average across datasets, significantly improving over prior editors.

cs.LG↗

arXiv2025

Maestro: Learning to Collaborate via Conditional Listwise Policy Optimization for Multi-Agent LLMs

Multi-agent systems (MAS) built on Large Language Models (LLMs) are being used to approach complex problems and can surpass single model inference. However, their success hinges on navigating a fundamental cognitive tension: the need to balance broad, divergent exploration of the solution space with a principled, convergent synthesis to the optimal solution. Existing paradigms often struggle to manage this duality, leading to premature consensus, error propagation, and a critical credit assignment problem that fails to distinguish between genuine reasoning and superficially plausible arguments. To resolve this core challenge, we propose the Multi-Agent Exploration-Synthesis framework Through Role Orchestration (Maestro), a principled paradigm for collaboration that structurally decouples these cognitive modes. Maestro uses a collective of parallel Execution Agents for diverse exploration and a specialized Central Agent for convergent, evaluative synthesis. To operationalize this critical synthesis phase, we introduce Conditional Listwise Policy Optimization (CLPO), a reinforcement learning objective that disentangles signals for strategic decisions and tactical rationales. By combining decision-focused policy gradients with a list-wise ranking loss over justifications, CLPO achieves clean credit assignment and stronger comparative supervision. Experiments on mathematical reasoning and general problem-solving benchmarks demonstrate that Maestro, coupled with CLPO, consistently outperforms existing state-of-the-art multi-agent approaches, delivering absolute accuracy gains of 6% on average and up to 10% at best.

cs.AI↗

arXiv2025

Nearly Instance-Optimal Parameter Recovery from Many Trajectories via Hellinger Localization

Learning from temporally-correlated data is a core facet of modern machine learning. Yet our understanding of sequential learning remains incomplete, particularly in the multi-trajectory setting where data consists of many independent realizations of a time-indexed stochastic process. This important regime both reflects modern training pipelines such as for large foundation models, and offers the potential for learning without the typical mixing assumptions made in the single-trajectory case. However, instance-optimal bounds are known only for least-squares regression with dependent covariates; for more general models or loss functions, the only broadly applicable guarantees result from a reduction to either i.i.d. learning, with effective sample size scaling only in the number of trajectories, or an existing single-trajectory result when each individual trajectory mixes, with effective sample size scaling as the full data budget deflated by the mixing-time. In this work, we significantly broaden the scope of instance-optimal rates in multi-trajectory settings via the Hellinger localization framework, a general approach for maximum likelihood estimation. Our method proceeds by first controlling the squared Hellinger distance at the path-measure level via a reduction to i.i.d. learning, followed by localization as a quadratic form in parameter space weighted by the trajectory Fisher information. This yields instance-optimal bounds that scale with the full data budget under a broad set of conditions. We instantiate our framework across four diverse case studies: a simple mixture of Markov chains, dependent linear regression under non-Gaussian noise, generalized linear models with non-monotonic activations, and linear-attention sequence models. In all cases, our bounds nearly match the instance-optimal rates from asymptotic normality, substantially improving over standard reductions.

cs.LG↗

arXiv2025

Latent Representations for Control Design with Provable Stability and Safety Guarantees

We initiate a formal study on the use of low-dimensional latent representations of dynamical systems for verifiable control synthesis. Our main goal is to enable the application of verification techniques -- such as Lyapunov or barrier functions -- that might otherwise be computationally prohibitive when applied directly to the full state representation. Towards this goal, we first provide dynamics-aware approximate conjugacy conditions which formalize the notion of reconstruction error necessary for systems analysis. We then utilize our conjugacy conditions to transfer the stability and invariance guarantees of a latent certificate function (e.g., a Lyapunov or barrier function) for a latent space controller back to the original system. Importantly, our analysis contains several important implications for learning latent spaces and dynamics, by highlighting the necessary geometric properties which need to be preserved by the latent space, in addition to providing concrete loss functions for dynamics reconstruction that are directly related to control design. We conclude by demonstrating the applicability of our theory to two case studies: (1) stabilization of a cartpole system, and (2) collision avoidance for a two vehicle system.

eess.SY↗

arXiv2025

CLAM: Continuous Latent Action Models for Robot Learning from Unlabeled Demonstrations

Learning robot control policies from demonstrations typically requires action-labeled expert data, which is expensive to collect through teleoperation. We study a more practical setting in which expert demonstrations are available only as observation sequences without action labels, and only task-agnostic play data contains actions. We introduce continuous latent action models (CLAM), a framework that infers continuous latent actions between consecutive observations using self-supervised dynamics prediction. To ground these latent actions into executable motor commands, CLAM jointly trains an action decoder using a small amount of task-agnostic play data. We show that continuous latent actions combined with this joint training are essential for high-dimensional continuous control. Across DMControl locomotion, MetaWorld manipulation, and real-world WidowX robot tasks, CLAM improves average task success rates by 2-3x over prior latent-action baselines and approaches behavior cloning trained with privileged expert action labels. Our results demonstrate that effective robot policies can be learned from unlabeled demonstrations and deployed on real hardware without collecting expert action-labeled data. Videos and code are available at clamrobot.github.io.

cs.RO↗

arXiv2025

Integration Matters for Learning PDEs with Backward SDEs

Backward stochastic differential equation (BSDE)-based deep learning methods provide an alternative to Physics-Informed Neural Networks (PINNs) for solving high-dimensional partial differential equations (PDEs), offering potential algorithmic advantages in settings such as stochastic optimal control, where the PDEs of interest are tied to an underlying dynamical system. However, standard BSDE-based solvers have empirically been shown to underperform relative to PINNs in the literature. In this paper, we identify the root cause of this performance gap as a discretization bias introduced by the standard Euler-Maruyama (EM) integration scheme applied to one-step self-consistency BSDE losses, which shifts the optimization landscape off target. We find that this bias cannot be satisfactorily addressed through finer step-sizes or multi-step self-consistency losses. To properly handle this issue, we propose a Stratonovich-based BSDE formulation, which we implement with stochastic Heun integration. We show that our proposed approach completely eliminates the bias issues faced by EM integration. Furthermore, our empirical results show that our Heun-based BSDE method consistently outperforms EM-based variants and achieves competitive results with PINNs across multiple high-dimensional benchmarks. Our findings highlight the critical role of integration schemes in BSDE-based PDE solvers, an algorithmic detail that has received little attention thus far in the literature.

cs.LG↗

arXiv2025

Data-Driven Reachability with Scenario Optimization and the Holdout Method

Reachability analysis is an important method in providing safety guarantees for systems with unknown or uncertain dynamics. Due to the computational intractability of exact reachability analysis for general nonlinear, high-dimensional systems, recent work has focused on the use of probabilistic methods for computing approximate reachable sets. In this work, we advocate for the use of a general purpose, practical, and sharp method for data-driven reachability: the holdout method. Despite the simplicity of the holdout method, we show -- on several numerical examples including scenario-based reach tubes -- that the resulting probabilistic bounds are substantially sharper and require fewer samples than existing methods for data-driven reachability. Furthermore, we complement our work with a discussion on the necessity of probabilistic reachability bounds. We argue that any method that attempts to de-randomize the bounds, by converting the guarantees to hold deterministically, requires (a) an exponential in state-dimension amount of samples to achieve non-vacuous guarantees, and (b) extra assumptions on the dynamics.

eess.SY↗

arXiv2024

Stability properties of gradient flow dynamics for the symmetric low-rank matrix factorization problem

The symmetric low-rank matrix factorization serves as a building block in many learning tasks, including matrix recovery and training of neural networks. However, despite a flurry of recent research, the dynamics of its training via non-convex factorized gradient-descent-type methods is not fully understood especially in the over-parameterized regime where the fitted rank is higher than the true rank of the target matrix. To overcome this challenge, we characterize equilibrium points of the gradient flow dynamics and examine their local and global stability properties. To facilitate a precise global analysis, we introduce a nonlinear change of variables that brings the dynamics into a cascade connection of three subsystems whose structure is simpler than the structure of the original system. We demonstrate that the Schur complement to a principal eigenspace of the target matrix is governed by an autonomous system that is decoupled from the rest of the dynamics. In the over-parameterized regime, we show that this Schur complement vanishes at an $O(1/t)$ rate, thereby capturing the slow dynamics that arises from excess parameters. We utilize a Lyapunov-based approach to establish exponential convergence of the other two subsystems. By decoupling the fast and slow parts of the dynamics, we offer new insight into the shape of the trajectories associated with local search algorithms and provide a complete characterization of the equilibrium points and their global stability properties. Such an analysis via nonlinear control techniques may prove useful in several related over-parameterized problems.

cs.LG↗

arXiv2024

Shallow diffusion networks provably learn hidden low-dimensional structure

Diffusion-based generative models provide a powerful framework for learning to sample from a complex target distribution. The remarkable empirical success of these models applied to high-dimensional signals, including images and video, stands in stark contrast to classical results highlighting the curse of dimensionality for distribution recovery. In this work, we take a step towards understanding this gap through a careful analysis of learning diffusion models over the Barron space of single layer neural networks. In particular, we show that these shallow models provably adapt to simple forms of low dimensional structure, thereby avoiding the curse of dimensionality. We combine our results with recent analyses of sampling with diffusion models to provide an end-to-end sample complexity bound for learning to sample from structured distributions. Importantly, our results do not require specialized architectures tailored to particular latent structures, and instead rely on the low-index structure of the Barron space to adapt to the underlying distribution.

cs.LG↗

arXiv2024

Incremental Composition of Learned Control Barrier Functions in Unknown Environments

We consider the problem of safely exploring a static and unknown environment while learning valid control barrier functions (CBFs) from sensor data. Existing works either assume known environments, target specific dynamics models, or use a-priori valid CBFs, and are thus limited in their safety guarantees for general systems during exploration. We present a method for safely exploring the unknown environment by incrementally composing a global CBF from locally-learned CBFs. The challenge here is that local CBFs may not have well-defined end behavior outside their training domain, i.e. local CBFs may be positive (indicating safety) in regions where no training data is available. We show that well-defined end behavior can be obtained when local CBFs are parameterized by compactly-supported radial basis functions. For learning local CBFs, we collect sensor data, e.g. LiDAR capturing obstacles in the environment, and augment it with simulated data from a safe oracle controller. Our work complements recent efforts to learn CBFs from safe demonstrations -- where learned safe sets are limited to their training domains -- by demonstrating how to grow the safe set over time as more data becomes available. We evaluate our approach on two simulated systems, where our method successfully explores an unknown environment while maintaining safety throughout the entire execution.

eess.SY↗

arXiv2024

Sharp Rates in Dependent Learning Theory: Avoiding Sample Size Deflation for the Square Loss

In this work, we study statistical learning with dependent ($\beta$-mixing) data and square loss in a hypothesis class $\mathscr{F}\subset L_{\Psi_p}$ where $\Psi_p$ is the norm $\|f\|_{\Psi_p} \triangleq \sup_{m\geq 1} m^{-1/p} \|f\|_{L^m} $ for some $p\in [2,\infty]$. Our inquiry is motivated by the search for a sharp noise interaction term, or variance proxy, in learning with dependent data. Absent any realizability assumption, typical non-asymptotic results exhibit variance proxies that are deflated multiplicatively by the mixing time of the underlying covariates process. We show that whenever the topologies of $L^2$ and $\Psi_p$ are comparable on our hypothesis class $\mathscr{F}$ -- that is, $\mathscr{F}$ is a weakly sub-Gaussian class: $\|f\|_{\Psi_p} \lesssim \|f\|_{L^2}^\eta$ for some $\eta\in (0,1]$ -- the empirical risk minimizer achieves a rate that only depends on the complexity of the class and second order statistics in its leading term. Our result holds whether the problem is realizable or not and we refer to this as a \emph{near mixing-free rate}, since direct dependence on mixing is relegated to an additive higher order term. We arrive at our result by combining the above notion of a weakly sub-Gaussian class with mixed tail generic chaining. This combination allows us to compute sharp, instance-optimal rates for a wide range of problems. Examples that satisfy our framework include sub-Gaussian linear regression, more general smoothly parameterized function classes, finite hypothesis classes, and bounded smoothness classes.

cs.LG↗

arXiv2023

Revisiting Energy Based Models as Policies: Ranking Noise Contrastive Estimation and Interpolating Energy Models

A crucial design decision for any robot learning pipeline is the choice of policy representation: what type of model should be used to generate the next set of robot actions? Owing to the inherent multi-modal nature of many robotic tasks, combined with the recent successes in generative modeling, researchers have turned to state-of-the-art probabilistic models such as diffusion models for policy representation. In this work, we revisit the choice of energy-based models (EBM) as a policy class. We show that the prevailing folklore -- that energy models in high dimensional continuous spaces are impractical to train -- is false. We develop a practical training objective and algorithm for energy models which combines several key ingredients: (i) ranking noise contrastive estimation (R-NCE), (ii) learnable negative samplers, and (iii) non-adversarial joint training. We prove that our proposed objective function is asymptotically consistent and quantify its limiting variance. On the other hand, we show that the Implicit Behavior Cloning (IBC) objective is actually biased even at the population level, providing a mathematical explanation for the poor performance of IBC trained energy policies in several independent follow-up works. We further extend our algorithm to learn a continuous stochastic process that bridges noise and data, modeling this process with a family of EBMs indexed by scale variable. In doing so, we demonstrate that the core idea behind recent progress in generative modeling is actually compatible with EBMs. Altogether, our proposed training algorithms enable us to train energy-based models as policies which compete with -- and even outperform -- diffusion models and other state-of-the-art approaches in several challenging multi-modal benchmarks: obstacle avoidance path planning and contact-rich block pushing.

cs.RO↗

arXiv2023

Robots That Ask For Help: Uncertainty Alignment for Large Language Model Planners

Large language models (LLMs) exhibit a wide range of promising capabilities -- from step-by-step planning to commonsense reasoning -- that may provide utility for robots, but remain prone to confidently hallucinated predictions. In this work, we present KnowNo, which is a framework for measuring and aligning the uncertainty of LLM-based planners such that they know when they don't know and ask for help when needed. KnowNo builds on the theory of conformal prediction to provide statistical guarantees on task completion while minimizing human help in complex multi-step planning settings. Experiments across a variety of simulated and real robot setups that involve tasks with different modes of ambiguity (e.g., from spatial to numeric uncertainties, from human preferences to Winograd schemas) show that KnowNo performs favorably over modern baselines (which may involve ensembles or extensive prompt tuning) in terms of improving efficiency and autonomy, while providing formal assurances. KnowNo can be used with LLMs out of the box without model-finetuning, and suggests a promising lightweight approach to modeling uncertainty that can complement and scale with the growing capabilities of foundation models. Website: https://robot-help.github.io

cs.RO↗

arXiv2023

Bootstrapped Representations in Reinforcement Learning

In reinforcement learning (RL), state representations are key to dealing with large or continuous state spaces. While one of the promises of deep learning algorithms is to automatically construct features well-tuned for the task they try to solve, such a representation might not emerge from end-to-end training of deep RL agents. To mitigate this issue, auxiliary objectives are often incorporated into the learning process and help shape the learnt state representation. Bootstrapping methods are today's method of choice to make these additional predictions. Yet, it is unclear which features these algorithms capture and how they relate to those from other auxiliary-task-based approaches. In this paper, we address this gap and provide a theoretical characterization of the state representation learnt by temporal difference learning (Sutton, 1988). Surprisingly, we find that this representation differs from the features learned by Monte Carlo and residual gradient algorithms for most transition structures of the environment in the policy evaluation setting. We describe the efficacy of these representations for policy evaluation, and use our theoretical analysis to design new auxiliary learning rules. We complement our theoretical results with an empirical comparison of these learning rules for different cumulant functions on classic domains such as the four-room domain (Sutton et al, 1999) and Mountain Car (Moore, 1990).

cs.LG↗