arXiv subjects

Baoxiang Wang

Publications and source records attributed to Baoxiang Wang.

At least 19 recordsLinked to original sources

arXiv2026

Learning Multi-Timescale Interventions under Safety and Resource Constraints

Many sequential decision problems offer qualitatively different ways of influencing the environment: some interventions act immediately, whereas others induce persistent effects that continue to shape future states long after the decision that initiated them. An agent must then decide jointly when to intervene, which temporal mode to use and how strongly, while accounting for residual effects and limited intervention resources. We introduce MINT: Multi-timescale Intervention Network Training. Persistent effects are carried by an augmented intervention state that accumulates and decays, while a structured policy separates intervention-mode selection from conditional control. Unlike temporal abstractions that extend policy execution, persistent-effect interventions remain part of the environment dynamics and may overlap with later interventions. We show that the augmented state is a sufficient statistic for the intervention history, preserving the Markov property, and establish Bellman contraction and almost-sure convergence of a tabular Q-learning instance. Across persistent-control MuJoCo locomotion, stochastic inventory management, and a physiologically grounded Type 1 Diabetes Mellitus (T1DM) simulator, MINT attains the best mean primary metric on two of three benchmarks while using fewer intervention activations than an identically augmented flat policy. Its return advantage is present at every binding intervention budget and narrows to parity in the unconstrained reference setting. In T1DM it achieves $90.9\pm0.9\%$ time in range with zero time below range, improving on the strongest baseline by $21.5\%$ points. Code, benchmarks and the configuration for every reported run are available at https://github.com/heywanrong/mint-rl.

cs.LG↗

arXiv2026

Impact of higher-order modes on eccentricity measurement in binary black hole gravitational waves

We investigate the systematic biases in measuring orbital eccentricity for binary black hole (BBH) mergers that arise when higher-order modes (HOMs) of gravitational waves are neglected in waveform modeling. Using Bayesian inference with the state-of-the-art eccentric, spin-aligned, higher-mode effective-one-body model SEOBNRv5EHM, we reanalyze six previously suggested eccentric gravitational-wave events--GW190521, GW190620, GW190701, GW191109, GW200129, and GW200208\_222617. Comparing results with its dominant-mode-only counterpart SEOBNRv5E, we find no statistically significant HOM-induced bias in eccentricity for any of these events, including GW190521, whose eccentricity has been debated in the literature. To explore representative parameter regimes vulnerable to HOM omission, we perform a broad zero-noise injection campaign varying detector-frame total mass, mass ratio, eccentricity, inclination, and network SNR. We find that significant systematic biases ($Δ_e/σ> 1$) arise predominantly in systems with relatively high total mass ($M^{\rm det}\gtrsim120M_\odot$), highly asymmetric mass ratios ($q \gtrsim 4$), large inclinations ($θ_\textrm{JN} \gtrsim 30^\circ$), and high SNRs ($ρ^N_\textrm{mf}\approx50$). Notably, for quasicircular BBHs with approximately $M^{\rm det}\gtrsim140M_\odot$, our results suggest that neglecting HOMs may lead to strong false-positive evidence for nonzero eccentricity. By contrast, for lower-mass systems ($M^{\rm det}\sim100 M_\odot$), HOM exclusion produces negligible eccentricity biases. Our results demonstrate that although current eccentric candidates are not impacted by HOM omission, future eccentricity measurements--particularly for relatively massive, asymmetric, or edge-on systems--require HOM-inclusive waveforms to avoid substantial systematic errors.

gr-qc↗

arXiv2026

Solver-Guided Reasoning for Mixed-Equilibrium Strategies

Reasoning in large language models (LLMs) is often grounded in human text, human demonstrations, and human-generated rationales. For equilibrium reasoning in complex games, however, relying on human data can be suboptimal. In fact, human play is often guided by intuition and heuristics and can deviate substantially from game equilibrium. This discrepancy is amplified in games with mixed-strategy equilibria, where human data is heavily biased toward pure strategies. Consequently, conditioning LLMs on this data yields weak game strategies. To grant LLMs the reasoning capacity in games, in this work, we study how to elicit equilibrium play using solver output. We propose Mixed-Strategy Decision Tree (MDT), which articulates the silent optimality of the equilibrium into sparse strategic rules that both humans and LLMs could understand. Using solver output rather than human annotation allows us to extend the input to arbitrarily new states and continuations. We instantiate this study on No-Limit Texas Hold'em by querying a solver oracle for over \textbf{250 million mixed-strategy decisions}; MDT together with other techniques \textbf{reduces the $\ell_1$ distance to the equilibrium by $52.6\%$} across $8$ different LLM configurations. A Route-only ablation tests the incremental contribution of the shadow-based contrast, while complete River-endgame and Liar's Dice experiments evaluate strategic fidelity and portability beyond the original NLH communication setting.

cs.LG↗

arXiv2026

Trust Region Masking for Long-Horizon LLM Reinforcement Learning

Policy gradient methods for Large Language Models optimize a policy $π_θ$ via a surrogate objective computed from samples of a rollout policy $π_{\text{roll}}$. However, modern LLM-RL pipelines suffer from unavoidable implementation divergences -- backend discrepancies, Mixture-of-Experts routing discontinuities, and distributed training staleness -- causing off-policy mismatch ($π_{\text{roll}} \neq π_θ$) and approximation errors between the surrogate and the true objective. We demonstrate that classical trust region bounds on this error scale as $O(T^2)$ with sequence length $T$, rendering them vacuous for long-horizon tasks. To address this, we derive a family of bounds -- both KL-based and TV-based -- including a Pinsker-Marginal bound ($O(T^{3/2})$), a Mixed bound ($O(T)$), and an Adaptive bound that strictly generalizes the Pinsker-Marginal bound via per-position importance-ratio decomposition. Taking the minimum over all bounds yields the tightest known guarantee across all divergence regimes. Crucially, all bounds depend on the maximum token-level divergence $D_{\mathrm{KL}}^{\mathrm{tok,max}}$ (or $D_{\mathrm{TV}}^{\mathrm{tok,max}}$), a sequence-level quantity that cannot be controlled by token-independent methods like PPO clipping. We propose Trust Region Masking (TRM), which masks entire sequences violating the trust region, enabling the first non-vacuous monotonic improvement guarantees for long-horizon LLM-RL.

cs.LG↗

arXiv2026

The Optimal Token Baseline: Variance Reduction for Long-Horizon LLM-RL

Reinforcement Learning (RL) for Large Language Models (LLMs) often suffers from training collapse in long-horizon tasks due to exploding gradient variance. To mitigate this, a baseline is commonly introduced for advantage computation; however, traditional value models remain difficult to optimize, and standard group-based baselines overlook sequence heterogeneity. Although classic optimal baseline theory can achieve global variance reduction, it neglects token heterogeneity and requires prohibitive gradient-based computation. In this work, we derive the Optimal Token Baseline (OTB) from first principles, proving that gradient updates should be weighted inversely to their cumulative gradient norm. To ensure efficiency, we propose the Logit-Gradient Proxy that approximates the gradient norm using only forward-pass probabilities. Our method achieves training stability and matches the performance of large group sizes ($N=32$) with only $N=4$, reducing token consumption by over 65\% across single-turn and tool-integrated reasoning tasks.

cs.LG↗

arXiv2026

Epistemic Gain, Aleatoric Cost: Uncertainty Decomposition in Multi-Agent Debate for Math Reasoning

Multi-Agent Debate (MAD) has shown promise in improving reasoning and reducing hallucinations, yet it remains unclear how information exchange shapes individual reasoning behavior. Empirically, MAD exhibits paradoxical phenomena, including rising accuracy with increasing token entropy and marked differences between homogeneous and heterogeneous agent combinations. In this paper, we introduce a Bayesian uncertainty analysis framework for MAD, which decomposes answer-level predictive uncertainty into epistemic uncertainty and aleatoric uncertainty, corresponding to the potential gain and cost of debate. Across multiple agent configurations, we find that effective debate depends on achieving high epistemic gain under controlled aleatoric cost. Building on this insight, we design an uncertainty-guided multi-agent reinforcement learning algorithm that encourages lower aleatoric cost and more effective epistemic information utilization. Experiments show that our approach simultaneously enhances each agent's accuracy and promotes a more productive debate process, providing an operational Bayesian perspective for understanding and improving MAD.

cs.MA↗

arXiv2026

Offline Multi-agent Reinforcement Learning via Sequential Score Decomposition

Offline cooperative multi-agent reinforcement learning (MARL) faces unique challenges due to distributional shifts, particularly stemming from the high dimensionality of joint action spaces and the presence of out-of-distribution joint action selections. In this work, we highlight that a fundamental challenge in offline MARL arises from the multi-equilibrium nature of cooperative tasks, which induces a highly multimodal joint behavior policy space coupled with heterogeneous-quality behavior data. This makes it difficult for individual policy regularization to align with a consistent coordination pattern, leading to the policy distribution shift problems. To tackle this challenge, we design a sequential score function decomposition method that distills per-agent regularization signals from the joint behavior policy, which induces coordinated modality selection under decentralized execution constraints. Then we leverage a flexible diffusion-based generative model to learn these score functions from multimodal offline data, and integrate them into joint-action critics to guide policy updates toward high-reward, in-distribution regions under a shared team reward. Our approach achieves state-of-the-art performance across multiple particle environments and Multi-agent MuJoCo benchmarks consistently. To the best of our knowledge, this is the first work to explicitly address the distributional gap between offline and online MARL, paving the way for more generalizable offline policy-based MARL methods.

cs.LG↗

arXiv2026

PokerSkill: LLMs Can Play Expert-Level Poker without Training or Solvers

Poker is a landmark challenge for artificial intelligence. The dominant approach relies on equilibrium solvers built on counterfactual regret minimization, requiring millions of core-hours of training. Large Language Models (LLMs) possess extensive poker knowledge but perform far below solver-based agents when asked to play directly. Traditional rule-based poker agents are interpretable and training-free, but their strategic ceiling remains far below equilibrium play. We introduce \textbf{PokerSkill}, a training-free and solver-free framework that bridges this gap by using detailed rule-based poker skills as a structured action-grounding interface for LLMs. A deterministic context engine analyzes the current state and retrieves only the relevant fragments from a layered skill library, which is entirely designed by human poker experts, constraining the LLM's choice to reasonable actions. Against GTOWizard, a state-of-the-art GTO benchmark, GPT-5.5 XHigh with PokerSkill achieves $-57 \pm 21$ mbb/hand, Claude Opus 4.6 achieves $-80 \pm 29$ mbb/hand and Claude Opus 4.7 achieves $-87\pm 64$ mbb/hand, reducing losses by 49--61\% compared to default-prompt baselines and outperforming the strong bot Slumbot. Our key finding is that rule-based skills alone do not constitute a strong strategy, and LLMs alone cannot play well, but their combination yields an agent that requires neither training nor solver access yet competes with systems built on millions of core-hours of computation. To our knowledge, this is the first demonstration of an LLM achieving competitive performance in a complex imperfect-information game without game-specific training or solver queries. Code is available at https://github.com/lbn187/PokerSkill.

cs.AI↗

arXiv2026

Talk, Judge, Cooperate: Gossip-Driven Indirect Reciprocity in Self-Interested LLM Agents

Indirect reciprocity, which means helping those who have helped others, is difficult to sustain among decentralized, self-interested LLM agents without reliable reputation systems. We address this challenge with the Agentic Linguistic Gossip Network (ALIGN), an automated framework that enables decentralized agents to form reputations, evaluate trustworthiness, and coordinate social norms by strategically sharing open-ended gossip with hierarchical tones. We demonstrate that ALIGN consistently improves indirect reciprocity and resists malicious entrants by identifying and ostracizing defectors. Notably, we find that stronger reasoning capabilities in LLMs lead to more incentive-aligned cooperation, whereas chat models often over-cooperate even when strategically suboptimal. These results suggest that leveraging LLM reasoning through decentralized gossip is a promising path for maintaining social welfare in agentic ecosystems. Our code is available at https://github.com/shuhui-zhu/ALIGN.

cs.MA↗

arXiv2026

The Reciprocity Gradient

Communication is fundamental to sustaining reciprocity and cooperation in strategic interactions. We identify and formulate the influence attribution problem as the central optimization difficulty inherent in such dynamics for a learning agent: any action or signal the agent emits reshapes the reputations of many third parties along combinatorially branching paths before feeding back into its own future rewards, forcing the agent to account for all of these indirect channels at once when choosing every action. To address this, we introduce the reciprocity gradient, which explicitly backpropagates reward gradients through private estimators of opponents' policies trained from public observations. The gradient flows through the reputation chain itself analytically, rather than being estimated from sampled returns. It jointly optimizes actions and evaluative signals without intrinsic rewards or reward shaping. Empirically, the method recovers near-optimal context-sensitive policies, while sample-based baselines collapse into constant-output policies.

cs.LG↗

arXiv2026

GW231123: False Massive Graviton Signatures from Unmodeled Point-Mass Lensing

GW231123 is the strongest current candidate for a lensed gravitational wave event and a unique case for testing how point-mass lensing affects propagation-based tests of gravity. In the real GW231123 data, an unlensed IMRPhenomXPHM analysis yields an apparent nonzero graviton mass posterior. We show that this anomaly is naturally explained by unmodeled point-mass lensing: once lensing is included, the apparent graviton mass signal disappears. In GW231123-like injection-recovery tests, a lensed NRSur7dq4 signal with zero graviton mass, recovered with the same unlensed IMRPhenomXPHM template, produces a similarly pronounced spurious graviton mass posterior, whereas lensing-included analyses with IMRPhenomXPHM, IMRPhenomXO4a, and NRSur7dq4 remain mutually consistent with no evidence for nonzero graviton mass. The similarity between the injected and real data posteriors shows that unmodeled point-mass lensing can mimic modified gravitational wave propagation. These results identify a concrete failure mode in tests of gravity and strengthen the interpretation of GW231123 as a lensed candidate.

gr-qc↗

arXiv2026

Dynamic Vocabulary Pruning: Stable LLM-RL by Taming the Tail

Reinforcement Learning (RL) for Large Language Models (LLMs) faces a fundamental tension: the numerical divergence between high-throughput inference engines and numerically precise training engines. Although these systems share the same parameters, they produce slightly different probability distributions, creating a training-inference mismatch. We prove that the bound on the log-probability divergence arising from this mismatch scales as $(1-p)$, where $p$ is the token probability. This scaling induces a highly asymmetric effect: the bound vanishes for high-probability tokens but remains significant for low-probability tokens in the distribution tail. When sampled, these tail tokens introduce systematically biased errors that accumulate over sequences, thereby destabilizing gradient estimation. Instead of applying post-hoc corrections, we propose Dynamic Vocabulary Pruning (DVP), which constrains the RL objective to a dynamically determined ''safe'' vocabulary that excludes the extreme tail. This strategy trades large, destabilizing numerical errors for a small, bounded optimization bias. We validate DVP empirically by demonstrating stable training, and theoretically by deriving strict bounds on the induced bias.

cs.LG↗

arXiv2025

Policy-Conditioned Policies for Multi-Agent Task Solving

In multi-agent tasks, the central challenge lies in the dynamic adaptation of strategies. However, directly conditioning on opponents' strategies is intractable in the prevalent deep reinforcement learning paradigm due to a fundamental ``representational bottleneck'': neural policies are opaque, high-dimensional parameter vectors that are incomprehensible to other agents. In this work, we propose a paradigm shift that bridges this gap by representing policies as human-interpretable source code and utilizing Large Language Models (LLMs) as approximate interpreters. This programmatic representation allows us to operationalize the game-theoretic concept of \textit{Program Equilibrium}. We reformulate the learning problem by utilizing LLMs to perform optimization directly in the space of programmatic policies. The LLM functions as a point-wise best-response operator that iteratively synthesizes and refines the ego agent's policy code to respond to the opponent's strategy. We formalize this process as \textit{Programmatic Iterated Best Response (PIBR)}, an algorithm where the policy code is optimized by textual gradients, using structured feedback derived from game utility and runtime unit tests. We demonstrate that this approach effectively solves several standard coordination matrix games and a cooperative Level-Based Foraging environment.

cs.GT↗

arXiv2025

Scalable Exploration via Ensemble++

Thompson Sampling is a principled method for balancing exploration and exploitation, but its real-world adoption faces computational challenges in large-scale or non-conjugate settings. While ensemble-based approaches offer partial remedies, they typically require prohibitively large ensemble sizes. We propose Ensemble++, a scalable exploration framework using a novel shared-factor ensemble architecture with random linear combinations. For linear bandits, we provide theoretical guarantees showing that Ensemble++ achieves regret comparable to exact Thompson Sampling with only $Θ(d \log T)$ ensemble sizes--significantly outperforming prior methods. Crucially, this efficiency holds across both compact and finite action sets with either time-invariant or time-varying contexts without configuration changes. We extend this theoretical foundation to nonlinear rewards by replacing fixed features with learnable neural representations while preserving the same incremental update principle, effectively bridging theory and practice for real-world tasks. Comprehensive experiments across linear, quadratic, neural, and GPT-based contextual bandits validate our theoretical findings and demonstrate Ensemble++'s superior regret-computation tradeoff versus state-of-the-art methods.

cs.LG↗

arXiv2025

Constraining scalar-tensor theories from higher harmonics with GW230529

The Advanced LIGO and Virgo collaborations recently detected a gravitational wave event, GW230529\_181500, during the fourth observing run, which is most plausibly attributed to the merger of a neutron star and a black hole. This observation provides an opportunity to test a class of gravitational theories that deviate from general relativity. In such theories, additional terms contribute to the gravitational wave signal only in cases of asymmetric binaries. This study focuses on two scalar-tensor models within this class of theories: the Screened Modified Gravity and Brans-Dicke theory. These models have potential applications in areas such as dark matter, dark energy, cosmic inflation, and primordial nucleosynthesis. With the GW230529\_181500 and Bayesian Markov-chain Monte Carlo analyses, we derive a 90\% credible lower bound as $ \frac{φ_{\rm_{VEV}}}{M_{\rm Pl}}<1.7\times10^{-2}$ and $ω_{\rm BD}>25.12$ by using dominant mode correction. Asymmetric binary systems usually have a significant mass ratio, in such cases, higher harmonic modes cannot be neglected. Our work considers higher harmonic corrections from scalar-tensor theories and provides a tighter constraint of $ \frac{φ_{\rm_{VEV}}}{M_{\rm Pl}}<1.5\times10^{-2}$ and $ω_{\rm BD}>32.68$, with a 13.3\% and 30.1\% improvement respectively. Combining GW230529\_181500, GW200115 and GW190814 and including higher modes, the constraint is improved to $\frac{φ_{\rm_{VEV}}}{M_{\rm Pl}}<7.84\times10^{-3}$ and $ω_{\rm BD}>123.75$. This is currently the strongest constraint from GWs, contingent upon GW190814 being an NSBH event.

gr-qc↗

arXiv2025

Information Bargaining: Bilateral Commitment in Bayesian Persuasion

Bayesian persuasion, an extension of cheap-talk communication, involves an informed sender committing to a signaling scheme to influence a receiver's actions. Compared to cheap talk, this sender's commitment enables the receiver to verify the incentive compatibility of signals beforehand, facilitating cooperation. While effective in one-shot scenarios, Bayesian persuasion faces computational complexity (NP-hardness) when extended to long-term interactions, where the receiver may adopt dynamic strategies conditional on past outcomes and future expectations. To address this complexity, we introduce the bargaining perspective, which allows: (1) a unified framework and well-structured solution concept for long-term persuasion, with desirable properties such as fairness and Pareto efficiency; (2) a clear distinction between two previously conflated advantages: the sender's informational advantage and first-proposer advantage. With only modest modifications to the standard setting, this perspective makes explicit the common knowledge of the game structure and grants the receiver comparable commitment capabilities, thereby reinterpreting classic one-sided persuasion as a balanced information bargaining framework. The framework is validated through a two-stage validation-and-inference paradigm: We first demonstrate that GPT-o3 and DeepSeek-R1, out of publicly available LLMs, reliably handle standard tasks; We then apply them to persuasion scenarios to test that the outcomes align with what our information-bargaining framework suggests. All code, results, and terminal logs are publicly available at github.com/YueLin301/InformationBargaining.

cs.GT↗

arXiv2025

Online Control with Adversarial Disturbance for Continuous-time Linear Systems

We study online control for continuous-time linear systems with finite sampling rates, where the objective is to design an online procedure that learns under non-stochastic noise and performs comparably to a fixed optimal linear controller. We present a novel two-level online algorithm, by integrating a higher-level learning strategy and a lower-level feedback control strategy. This method offers a practical and robust solution for online control, which achieves sublinear regret. Our work provides the first nonasymptotic results for controlling continuous-time linear systems with finite number of interactions with the system. Moreover, we examine how to train an agent in domain randomization environments from a non-stochastic control perspective. By applying our method to the SAC (Soft Actor-Critic) algorithm, we achieved improved results in multiple reinforcement learning tasks within domain randomization environments. Our work provides new insights into non-asymptotic analyses of controlling continuous-time systems. Furthermore, our work brings practical intuition into controller learning under non-stochastic environments.

math.OC↗

arXiv2025

ADG: Ambient Diffusion-Guided Dataset Recovery for Corruption-Robust Offline Reinforcement Learning

Real-world datasets collected from sensors or human inputs are prone to noise and errors, posing significant challenges for applying offline reinforcement learning (RL). While existing methods have made progress in addressing corrupted actions and rewards, they remain insufficient for handling corruption in high-dimensional state spaces and for cases where multiple elements in the dataset are corrupted simultaneously. Diffusion models, known for their strong denoising capabilities, offer a promising direction for this problem-but their tendency to overfit noisy samples limits their direct applicability. To overcome this, we propose Ambient Diffusion-Guided Dataset Recovery (ADG), a novel approach that pioneers the use of diffusion models to tackle data corruption in offline RL. First, we introduce Ambient Denoising Diffusion Probabilistic Models (DDPM) from approximated distributions, which enable learning on partially corrupted datasets with theoretical guarantees. Second, we use the noise-prediction property of Ambient DDPM to distinguish between clean and corrupted data, and then use the clean subset to train a standard DDPM. Third, we employ the trained standard DDPM to refine the previously identified corrupted data, enhancing data quality for subsequent offline RL training. A notable strength of ADG is its versatility-it can be seamlessly integrated with any offline RL algorithm. Experiments on a range of benchmarks, including MuJoCo, Kitchen, and Adroit, demonstrate that ADG effectively mitigates the impact of corrupted data and improves the robustness of offline RL under various noise settings, achieving state-of-the-art results.

cs.LG↗