Haimin Hu

Yiyou Sun, Xinyang Han, Weichen Zhang et al.

Recent AI systems have achieved strong results on a wide range of benchmarks, yet these gains have not translated into economically meaningful deployment across many professional domains. We argue that this gap is largely an evaluation problem: widely used benchmarks lack sustained performance measurement on real and economically valuable workflows. This paper introduces Agents' Last Exam (ALE), a benchmark designed to evaluate AI agents on long-horizon, economically valuable, real-world tasks with verifiable outcomes. Developed in collaboration with 250+ industry experts, ALE covers non-physical industries defined with reference to O*NET / SOC 2018 (the U.S. federal occupational taxonomy). It is organized around a task taxonomy with 55 subfields grouped into 13 industry clusters covering 1K+ tasks. Current results show that the hardest tier remains far from saturated: across mainstream harness and backbone configurations, the average full pass rate is 2.6%. ALE is designed as a living benchmark: its task pool grows continuously as new workflows and industries are onboarded. More broadly, ALE is intended not merely as another leaderboard, but as an instrument for closing the gap between benchmark success and GDP-relevant impact.

Haimin Hu

Autonomous robots that interact with people must make safe and efficient decisions under human-induced uncertainty, such as their preferences, goals, competency, and willingness to cooperate. Safety filters are a popular approach for ensuring safety in interactive robotics, since their modular design separates safety from performance, allowing robots to operate safely around people with minimal impact on task efficiency. While traditional safety filters typically operate only in the physical space, neglecting the robot's ability to learn and adapt online, the recently proposed belief-space safety filter (BeliefSF) reasons about robot safety in closed-loop with runtime inference that actively reduces the robot's uncertainty online, thereby reducing conservativeness in filtering. However, providing formal safety guarantees for robots deploying BeliefSF remains a significant challenge due to errors in runtime inference and neural approximation of safety filters required to handle the high dimensionality of belief spaces. In this paper, we propose an algorithmic approach to certify high-probability safety of BeliefSF using conformal prediction, while explicitly accounting for the reliability of the robot's runtime inference module. Our method leverages the structure of belief-space safety filtering by focusing verification on a region where inference is expected to be reliable. It preserves the simplicity and sample complexity of standard conformal prediction, yet can certify a substantially less conservative safety filter. Through a simulated human-vehicle interaction benchmark, we show that our approach verifies a significantly more permissive belief-space safety filter than a standard conformal prediction baseline.

Kai-Chieh Hsu, Haimin Hu, Jaime Fernández Fisac

Recent years have seen significant progress in the realm of robot autonomy, accompanied by the expanding reach of robotic technologies. However, the emergence of new deployment domains brings unprecedented challenges in ensuring safe operation of these systems, which remains as crucial as ever. While traditional model-based safe control methods struggle with generalizability and scalability, emerging data-driven approaches tend to lack well-understood guarantees, which can result in unpredictable catastrophic failures. Successful deployment of the next generation of autonomous robots will require integrating the strengths of both paradigms. This article provides a review of safety filter approaches, highlighting important connections between existing techniques and proposing a unified technical framework to understand, compare, and combine them. The new unified view exposes a shared modular structure across a range of seemingly disparate safety filter classes and naturally suggests directions for future progress towards more scalable synthesis, robust monitoring, and efficient intervention.

Haimin Hu, David Isele, Sangjae Bae et al.

The ability to accurately predict others' behavior is central to the safety and efficiency of interactive robotics. Unfortunately, robots often lack access to key information on which these predictions may hinge, such as other agents' goals, attention, and willingness to cooperate. Dual control theory addresses this challenge by treating unknown parameters of a predictive model as stochastic hidden states and inferring their values at runtime using information gathered during system operation. While able to optimally and automatically trade off exploration and exploitation, dual control is computationally intractable for general interactive motion planning. In this paper, we present a novel algorithmic approach to enable active uncertainty reduction for interactive motion planning based on the implicit dual control paradigm. Our approach relies on sampling-based approximation of stochastic dynamic programming, leading to a model predictive control problem that can be readily solved by real-time gradient-based optimization methods. The resulting policy is shown to preserve the dual control effect for a broad class of predictive models with both continuous and categorical uncertainty. To ensure the safe operation of the interacting agents, we use a runtime safety filter (also referred to as a "shielding" scheme), which overrides the robot's dual control policy with a safety fallback strategy when a safety-critical event is imminent. We then augment the dual control framework with an improved variant of the recently proposed shielding-aware robust planning scheme, which proactively balances the nominal planning performance with the risk of high-cost emergency maneuvers triggered by low-probability agent behaviors. We demonstrate the efficacy of our approach with both simulated driving studies and hardware experiments using 1/10 scale autonomous vehicles.

Justin Wang, Haimin Hu, Duy Phuong Nguyen et al.

While robust optimal control theory provides a rigorous framework to compute robot control policies that are provably safe, it struggles to scale to high-dimensional problems, leading to increased use of deep learning for tractable synthesis of robot safety. Unfortunately, existing neural safety synthesis methods often lack convergence guarantees and solution interpretability. In this paper, we present Minimax Actors Guided by Implicit Critic Stackelberg (MAGICS), a novel adversarial reinforcement learning (RL) algorithm that guarantees local convergence to a minimax equilibrium solution. We then build on this approach to provide local convergence guarantees for a general deep RL-based robot safety synthesis algorithm. Through both simulation studies on OpenAI Gym environments and hardware experiments with a 36-dimensional quadruped robot, we show that MAGICS can yield robust control policies outperforming the state-of-the-art neural safety synthesis methods.

Donggeon David Oh, Duy P. Nguyen, Haimin Hu et al.

Robust control barrier functions (CBFs) provide a principled mechanism for smooth safety enforcement under worst-case disturbances. However, existing approaches typically rely on explicit, closed-form structure in the dynamics (e.g., control-affine) and uncertainty models. This has led to limited scalability and generality, with most robust CBFs certifying only conservative subsets of the maximal robust safe set. In this paper, we introduce a new robust CBF framework for general nonlinear systems under bounded uncertainty. We first show that the safety value function solving the dynamic programming Isaacs equation is a valid robust discrete-time CBF that enforces safety on the maximal robust safe set. We then adopt the key reinforcement learning (RL) notion of quality function (or Q-function), which removes the need for explicit dynamics by lifting the barrier certificate into state-action space and yields a novel robust Q-CBF constraint for safety filtering. Combined with adversarial RL, this enables the synthesis and deployment of robust Q-CBFs on general nonlinear systems with black-box dynamics and unknown uncertainty structure. We validate the framework on a canonical inverted pendulum benchmark and a 36-D quadruped simulator, achieving substantially less conservative safe sets than barrier-based baselines on the pendulum and reliable safety enforcement even under adversarial uncertainty realizations on the quadruped.

Kaiqu Liang, Haimin Hu, Ryan Liu et al. · princeton

While Reinforcement Learning from Human Feedback (RLHF) has shown promise in aligning generative AI, we present empirical evidence that it can also cause severe, systematic misalignment. We hypothesize that this stems from evaluator feedback depending on downstream outcome predictions (foresight) that can be influenced by the AI's output, inducing Goodhart's law dynamics. We present a theoretical analysis showing that conditioning evaluator feedback on downstream observations (hindsight) inhibits this effect by decoupling the alignment signal from potentially compromised predictions--crucially, the result holds even if the observed outcomes are sampled from the AI's own world model. Building on this insight, we introduce Reinforcement Learning from Hindsight Simulation (RLHS), which presents plausible simulated outcomes to evaluators before eliciting feedback. We validate RLHS across three consultancy settings--marketplace interactions, restaurant recommendations, and online course advising--using both online (PPO) and offline (DPO) fine-tuning methods, and show that it substantially improves alignment over RLHF in experiments and human evaluations. We perform post-hoc benchmark evaluations on TruthfulQA, HaluEval, and TrustLLM, finding that even after single-task fine-tuning, RLHF misalignment persists, whereas RLHS consistently outperforms baselines and demonstrates robust alignment generalization. The project webpage and code are available at https://rl-hindsight.github.io.

Justin Lidard, Haimin Hu, Asher Hancock et al.

As intelligent robots like autonomous vehicles become increasingly deployed in the presence of people, the extent to which these systems should leverage model-based game-theoretic planners versus data-driven policies for safe, interaction-aware motion planning remains an open question. Existing dynamic game formulations assume all agents are task-driven and behave optimally. However, in reality, humans tend to deviate from the decisions prescribed by these models, and their behavior is better approximated under a noisy-rational paradigm. In this work, we investigate a principled methodology to blend a data-driven reference policy with an optimization-based game-theoretic policy. We formulate KLGame, an algorithm for solving non-cooperative dynamic game with Kullback-Leibler (KL) regularization with respect to a general, stochastic, and possibly multi-modal reference policy. Our method incorporates, for each decision maker, a tunable parameter that permits modulation between task-driven and data-driven behaviors. We propose an efficient algorithm for computing multi-modal approximate feedback Nash equilibrium strategies of KLGame in real time. Through a series of simulated and real-world autonomous driving scenarios, we demonstrate that KLGame policies can more effectively incorporate guidance from the reference policy and account for noisily-rational human behaviors versus non-regularized baselines. Website with additional information, videos, and code: https://kl-games.github.io/.

Haimin Hu, Gabriele Dragotto, Zixu Zhang et al. · princeton

We consider the multi-agent spatial navigation problem of computing the socially optimal order of play, i.e., the sequence in which the agents commit to their decisions, and its associated equilibrium in an N-player Stackelberg trajectory game. We model this problem as a mixed-integer optimization problem over the space of all possible Stackelberg games associated with the order of play's permutations. To solve the problem, we introduce Branch and Play (B&P), an efficient and exact algorithm that provably converges to a socially optimal order of play and its Stackelberg equilibrium. As a subroutine for B&P, we employ and extend sequential trajectory planning, i.e., a popular multi-agent control approach, to scalably compute valid local Stackelberg equilibria for any given order of play. We demonstrate the practical utility of B&P to coordinate air traffic control, swarm formation, and delivery vehicle fleets. We find that B&P consistently outperforms various baselines, and computes the socially optimal equilibrium.

Donggeon David Oh, Duy P. Nguyen, Haimin Hu et al.

Recent advances in reinforcement learning (RL) enable its use on increasingly complex tasks, but the lack of formal safety guarantees still limits its application in safety-critical settings. A common practical approach is to augment the RL policy with a safety filter that overrides unsafe actions to prevent failures during both training and deployment. However, safety filtering is often perceived as sacrificing performance and hindering the learning process. We show that this perceived safety-performance tradeoff is not inherent and prove, for the first time, that enforcing safety with a sufficiently permissive safety filter does not degrade asymptotic performance. We formalize RL safety with a safety-critical Markov decision process (SC-MDP), which requires categorical, rather than high-probability, avoidance of catastrophic failure states. Additionally, we define an associated filtered MDP in which all actions result in safe effects, thanks to a safety filter that is considered to be a part of the environment. Our main theorem establishes that (i) learning in the filtered MDP is safe categorically, (ii) standard RL convergence carries over to the filtered MDP, and (iii) any policy that is optimal in the filtered MDP-when executed through the same filter-achieves the same asymptotic return as the best safe policy in the SC-MDP, yielding a complete separation between safety enforcement and performance optimization. We validate the theory on Safety Gymnasium with representative tasks and constraints, observing zero violations during training and final performance matching or exceeding unfiltered baselines. Together, these results shed light on a long-standing question in safety-filtered learning and provide a simple, principled recipe for safe RL: train and deploy RL policies with the most permissive safety filter that is available.

Kaiqu Liang, Haimin Hu, Xuandong Zhao et al. · berkeley, princeton

Bullshit, as conceptualized by philosopher Harry Frankfurt, refers to statements made without regard to their truth value. While previous work has explored large language model (LLM) hallucination and sycophancy, we propose machine bullshit as an overarching conceptual framework that can allow researchers to characterize the broader phenomenon of emergent loss of truthfulness in LLMs and shed light on its underlying mechanisms. We introduce the Bullshit Index, a novel metric quantifying LLMs' indifference to truth, and propose a complementary taxonomy analyzing four qualitative forms of bullshit: empty rhetoric, paltering, weasel words, and unverified claims. We conduct empirical evaluations on the Marketplace dataset, the Political Neutrality dataset, and our new BullshitEval benchmark (2,400 scenarios spanning 100 AI assistants) explicitly designed to evaluate machine bullshit. Our results demonstrate that model fine-tuning with reinforcement learning from human feedback (RLHF) significantly exacerbates bullshit and inference-time chain-of-thought (CoT) prompting notably amplify specific bullshit forms, particularly empty rhetoric and paltering. We also observe prevalent machine bullshit in political contexts, with weasel words as the dominant strategy. Our findings highlight systematic challenges in AI alignment and provide new insights toward more truthful LLM behavior.

Haimin Hu, Zixu Zhang, Kensuke Nakamura et al.

An outstanding challenge for the widespread deployment of robotic systems like autonomous vehicles is ensuring safe interaction with humans without sacrificing performance. Existing safety methods often neglect the robot's ability to learn and adapt at runtime, leading to overly conservative behavior. This paper proposes a new closed-loop paradigm for synthesizing safe control policies that explicitly account for the robot's evolving uncertainty and its ability to quickly respond to future scenarios as they arise, by jointly considering the physical dynamics and the robot's learning algorithm. We leverage adversarial reinforcement learning for tractable safety analysis under high-dimensional learning dynamics and demonstrate our framework's ability to work with both Bayesian belief propagation and implicit learning through large pre-trained neural trajectory predictors.

Haimin Hu, Jaime F. Fisac

The ability to accurately predict human behavior is central to the safety and efficiency of robot autonomy in interactive settings. Unfortunately, robots often lack access to key information on which these predictions may hinge, such as people's goals, attention, and willingness to cooperate. Dual control theory addresses this challenge by treating unknown parameters of a predictive model as stochastic hidden states and inferring their values at runtime using information gathered during system operation. While able to optimally and automatically trade off exploration and exploitation, dual control is computationally intractable for general interactive motion planning, mainly due to the fundamental coupling between robot trajectory optimization and human intent inference. In this paper, we present a novel algorithmic approach to enable active uncertainty reduction for interactive motion planning based on the implicit dual control paradigm. Our approach relies on sampling-based approximation of stochastic dynamic programming, leading to a model predictive control problem that can be readily solved by real-time gradient-based optimization methods. The resulting policy is shown to preserve the dual control effect for a broad class of predictive human models with both continuous and categorical uncertainty. The efficacy of our approach is demonstrated with simulated driving examples.

Haimin Hu, Kensuke Nakamura, Jaime F. Fisac

Jointly achieving safety and efficiency in human-robot interaction (HRI) settings is a challenging problem, as the robot's planning objectives may be at odds with the human's own intent and expectations. Recent approaches ensure safe robot operation in uncertain environments through a supervisory control scheme, sometimes called "shielding", which overrides the robot's nominal plan with a safety fallback strategy when a safety-critical event is imminent. These reactive "last-resort" strategies (typically in the form of aggressive emergency maneuvers) focus on preserving safety without efficiency considerations; when the nominal planner is unaware of possible safety overrides, shielding can be activated more frequently than necessary, leading to degraded performance. In this work, we propose a new shielding-based planning approach that allows the robot to plan efficiently by explicitly accounting for possible future shielding events. Leveraging recent work on Bayesian human motion prediction, the resulting robot policy proactively balances nominal performance with the risk of high-cost emergency maneuvers triggered by low-probability human behaviors. We formalize Shielding-Aware Robust Planning (SHARP) as a stochastic optimal control problem and propose a computationally efficient framework for finding tractable approximate solutions at runtime. Our method outperforms the shielding-agnostic motion planning baseline (equipped with the same human intent inference scheme) on simulated driving examples with human trajectories taken from the recently released Waymo Open Motion Dataset.

Mo Chen, Sylvia L. Herbert, Haimin Hu et al.

Real-time, guaranteed safe trajectory planning is vital for navigation in unknown environments. However, real-time navigation algorithms typically sacrifice robustness for computation speed. Alternatively, provably safe trajectory planning tends to be too computationally intensive for real-time replanning. We propose FaSTrack, Fast and Safe Tracking, a framework that achieves both real-time replanning and guaranteed safety. In this framework, real-time computation is achieved by allowing any trajectory planner to use a simplified \textit{planning model} of the system. The plan is tracked by the system, represented by a more realistic, higher-dimensional \textit{tracking model}. We precompute the tracking error bound (TEB) due to mismatch between the two models and due to external disturbances. We also obtain the corresponding tracking controller used to stay within the TEB. The precomputation does not require prior knowledge of the environment. We demonstrate FaSTrack using Hamilton-Jacobi reachability for precomputation and three different real-time trajectory planners with three different tracking-planning model pairs.

Lars Lindemann, Haimin Hu, Alexander Robey et al.

Motivated by the lack of systematic tools to obtain safe control laws for hybrid systems, we propose an optimization-based framework for learning certifiably safe control laws from data. In particular, we assume a setting in which the system dynamics are known and in which data exhibiting safe system behavior is available. We propose hybrid control barrier functions for hybrid systems as a means to synthesize safe control inputs. Based on this notion, we present an optimization-based framework to learn such hybrid control barrier functions from data. Importantly, we identify sufficient conditions on the data such that feasibility of the optimization problem ensures correctness of the learned hybrid control barrier functions, and hence the safety of the system. We illustrate our findings in two simulations studies, including a compass gait walker.

Haimin Hu, Mahyar Fazlyab, Manfred Morari et al.

There has been an increasing interest in using neural networks in closed-loop control systems to improve performance and reduce computational costs for on-line implementation. However, providing safety and stability guarantees for these systems is challenging due to the nonlinear and compositional structure of neural networks. In this paper, we propose a novel forward reachability analysis method for the safety verification of linear time-varying systems with neural networks in feedback interconnection. Our technical approach relies on abstracting the nonlinear activation functions by quadratic constraints, which leads to an outer-approximation of forward reachable sets of the closed-loop system. We show that we can compute these approximate reachable sets using semidefinite programming. We illustrate our method in a quadrotor example, in which we first approximate a nonlinear model predictive controller via a deep neural network and then apply our analysis tool to certify finite-time reachability and constraint satisfaction of the closed-loop system.

Alexander Robey, Haimin Hu, Lars Lindemann et al.

Inspired by the success of imitation and inverse reinforcement learning in replicating expert behavior through optimal control, we propose a learning based approach to safe controller synthesis based on control barrier functions (CBFs). We consider the setting of a known nonlinear control affine dynamical system and assume that we have access to safe trajectories generated by an expert - a practical example of such a setting would be a kinematic model of a self-driving vehicle with safe trajectories (e.g., trajectories that avoid collisions with obstacles in the environment) generated by a human driver. We then propose and analyze an optimization-based approach to learning a CBF that enjoys provable safety guarantees under suitable Lipschitz smoothness assumptions on the underlying dynamical system. A strength of our approach is that it is agnostic to the parameterization used to represent the CBF, assuming only that the Lipschitz constant of such functions can be efficiently bounded. Furthermore, if the CBF parameterization is convex, then under mild assumptions, so is our learning process. We end with extensive numerical evaluations of our results on both planar and realistic examples, using both random feature and deep neural network parameterizations of the CBF. To the best of our knowledge, these are the first results that learn provably safe control barrier functions from data.