Search

Since the release of ChatGPT, large language models (LLMs) have demonstrated remarkable capabilities across various domains. A key challenge in developing these general capabilities is efficiently sourcing diverse, high-quality data. This becomes especially critical in reasoning-related tasks with sandbox checkers, such as math or code, where the goal is to generate correct solutions to specific problems with higher probability. In this work, we introduce Flaming-hot Initiation with Regular Execution (FIRE) sampling, a simple yet highly effective method to efficiently find good responses. Our empirical findings show that FIRE sampling enhances inference-time generation quality and also benefits training in the alignment stage. Furthermore, we explore how FIRE sampling improves performance by promoting diversity and analyze the impact of employing FIRE at different positions within a response.

Reinforcement Learning (RL) with unit test feedback has enhanced large language models (LLMs) code generation, but relies on sparse rewards provided only after complete code evaluation, limiting learning efficiency and incremental improvements. When generated code fails all unit tests, no learning signal is received, hindering progress on complex tasks. To address this, we propose a Process Reward Model (PRM) that delivers dense, line-level feedback on code correctness during generation, mimicking human code refinement and providing immediate guidance. We explore various strategies for training PRMs and integrating them into the RL framework, finding that using PRMs both as dense rewards and for value function initialization significantly boosts performance. Our approach increases our in-house LLM's pass rate from 28.2% to 29.8% on LiveCodeBench and from 31.8% to 35.8% on our internal benchmark. Our experimental results highlight the effectiveness of PRMs in enhancing RL-driven code generation, especially for long-horizon scenarios., Comment: 14 pages, 5 figures

Reinforcement Learning (RL) plays a crucial role in aligning large language models (LLMs) with human preferences and improving their ability to perform complex tasks. However, current approaches either require significant computational resources due to the use of multiple models and extensive online sampling for training (e.g., PPO) or are framed as bandit problems (e.g., DPO, DRO), which often struggle with multi-step reasoning tasks, such as math problem-solving and complex reasoning that involve long chains of thought. To overcome these limitations, we introduce Direct Q-function Optimization (DQO), which formulates the response generation process as a Markov Decision Process (MDP) and utilizes the soft actor-critic (SAC) framework to optimize a Q-function directly parameterized by the language model. The MDP formulation of DQO offers structural advantages over bandit-based methods, enabling more effective process supervision. Experimental results on two math problem-solving datasets, GSM8K and MATH, demonstrate that DQO outperforms previous methods, establishing it as a promising offline reinforcement learning approach for aligning language models.

Recent measurements of baryon acoustic oscillations (BAO) by the Dark Energy Spectroscopic Instrument (DESI) suggest a preference for a dynamic dark energy model over a cosmological constant. This conclusion emerges from the combination of DESI's BAO data with observations of the Cosmic Microwave Background (CMB) and various type Ia supernova (SN Ia) catalogues. The deviation observed in the cosmological constant ($\Lambda$) reflects a departure from the standard cosmological model. Testing this deviation involves examining the consistency between cosmological parameters derived from early and late-time observations. Specifically, we focus on the matter density parameter $\omega_m = \Omega_mh^2$ and introduce ${\rm ratio}(\omega_m)$ to assess consistency, which is defined as the ratio of $\omega_m$ values constrained by high and low-redshift measurements. This ratio serves as a metric for quantifying deviations from the $\Lambda$CDM model. In this paper, we find that the DESI BAO+CMB yields ${\rm ratio}(\omega_m)=1.0171\pm0.0066$. Upon excluding the LRG1 and LRG2 data in DESI BAO, this ratio adjusts to ${\rm ratio}(\omega_m)=1.0100\pm0.0082$. This shift, corresponding to a change from $2.6\sigma$ to $1.2\sigma$, indicates that the deviation from the $\Lambda$CDM model is predominantly driven by these two samples from the DESI BAO measurements. To substantiate this conclusion, we utilized two cosmological model-independent methods to reconstruct the cosmic expansion history. Both reconstructions of the Hubble parameter $H(z)$ indicate that the evolving features of dark energy are determined by the combined LRG1 and LRG2 data. Therefore, different methods have reached the same conclusion, namely the importance of accurately measuring the BAO feature in LRG1 and LRG2 data., Comment: 12 pages, 6 figures

The multi-agent reinforcement learning systems (MARL) based on the Markov decision process (MDP) have emerged in many critical applications. To improve the robustness/defense of MARL systems against adversarial attacks, the study of various adversarial attacks on reinforcement learning systems is very important. Previous works on adversarial attacks considered some possible features to attack in MDP, such as the action poisoning attacks, the reward poisoning attacks, and the state perception attacks. In this paper, we propose a brand-new form of attack called the camouflage attack in the MARL systems. In the camouflage attack, the attackers change the appearances of some objects without changing the actual objects themselves; and the camouflaged appearances may look the same to all the targeted recipient (victim) agents. The camouflaged appearances can mislead the recipient agents to misguided actions. We design algorithms that give the optimal camouflage attacks minimizing the rewards of recipient agents. Our numerical and theoretical results show that camouflage attacks can rival the more conventional, but likely more difficult state perception attacks. We also investigate cost-constrained camouflage attacks and showed numerically how cost budgets affect the attack performance., Comment: arXiv admin note: text overlap with arXiv:2311.00859

The recent local measurements of the Hubble constant $H_0$, indicate a significant discrepancy of over 5$\sigma$ compared to the value inferred from \textit{Planck} observations of the cosmic microwave background (CMB). In this paper, we try to understand the origin of this tension by testing the consistency of early and late cosmological parameters in the same observed data. In practice, we simultaneously derive the early and late parameters using baryon acoustic oscillation (BAO) measurements, which provide both low and high-redshift information. To resolve parameter degeneracy, the complementary data from CMB observations are included in the analysis. By using the parameter $\omega_m = \Omega_mh^2$, we introduce ${\rm ratio}(\omega_m)$, defined as the ratio of $\omega_m$ which are constrained from high and low-redshift measurements respectively, to quantify the consistency between early and late parameters. We obtained a value of ${\rm ratio}(\omega_m) = 1.0069\pm0.0070$, indicating there is no tension between early parameters and late parameters in the framework of $\Lambda$CDM model. As a result, the Hubble tension may arise from the differences of datasets or unknown systematic errors in the current data. In addition, we forecast the future BAO measurements of ${\rm ratio}(\omega_m)$, using several galaxy redshift surveys and 21 cm intensity mapping surveys, and find that these measurements can significantly improve the precision of cosmological parameters., Comment: 12 pages, 5 figures

Finding optimal adversarial attack strategies is an important topic in reinforcement learning and the Markov decision process. Previous studies usually assume one all-knowing coordinator (attacker) for whom attacking different recipient (victim) agents incurs uniform costs. However, in reality, instead of using one limitless central attacker, the attacks often need to be performed by distributed attack agents. We formulate the problem of performing optimal adversarial agent-to-agent attacks using distributed attack agents, in which we impose distinct cost constraints on each different attacker-victim pair. We propose an optimal method integrating within-step static constrained attack-resource allocation optimization and between-step dynamic programming to achieve the optimal adversarial attack in a multi-agent system. Our numerical results show that the proposed attacks can significantly reduce the rewards received by the attacked agents., Comment: Submitted to ICCASP2024

Due to the broad range of applications of multi-agent reinforcement learning (MARL), understanding the effects of adversarial attacks against MARL model is essential for the safe applications of this model. Motivated by this, we investigate the impact of adversarial attacks on MARL. In the considered setup, there is an exogenous attacker who is able to modify the rewards before the agents receive them or manipulate the actions before the environment receives them. The attacker aims to guide each agent into a target policy or maximize the cumulative rewards under some specific reward function chosen by the attacker, while minimizing the amount of manipulation on feedback and action. We first show the limitations of the action poisoning only attacks and the reward poisoning only attacks. We then introduce a mixed attack strategy with both the action poisoning and the reward poisoning. We show that the mixed attack strategy can efficiently attack MARL agents even if the attacker has no prior information about the underlying environment and the agents' algorithms.

Robust reinforcement learning (RL) aims to find a policy that optimizes the worst-case performance in the face of uncertainties. In this paper, we focus on action robust RL with the probabilistic policy execution uncertainty, in which, instead of always carrying out the action specified by the policy, the agent will take the action specified by the policy with probability $1-\rho$ and an alternative adversarial action with probability $\rho$. We establish the existence of an optimal policy on the action robust MDPs with probabilistic policy execution uncertainty and provide the action robust Bellman optimality equation for its solution. Furthermore, we develop Action Robust Reinforcement Learning with Certificates (ARRLC) algorithm that achieves minimax optimal regret and sample complexity. Furthermore, we conduct numerical experiments to validate our approach's robustness, demonstrating that ARRLC outperforms non-robust RL algorithms and converges faster than the robust TD algorithm in the presence of action perturbations.

Contextual bandit algorithms have many applicants in a variety of scenarios. In order to develop trustworthy contextual bandit systems, understanding the impacts of various adversarial attacks on contextual bandit algorithms is essential. In this paper, we propose a new class of attacks: action poisoning attacks, where an adversary can change the action signal selected by the agent. We design action poisoning attack schemes against linear contextual bandit algorithms in both white-box and black-box settings. We further analyze the cost of the proposed attack strategies for a very popular and widely used bandit algorithm: LinUCB. We show that, in both white-box and black-box settings, the proposed attack schemes can force the LinUCB agent to pull a target arm very frequently by spending only logarithm cost.

Due to the broad range of applications of reinforcement learning (RL), understanding the effects of adversarial attacks against RL model is essential for the safe applications of this model. Prior theoretical works on adversarial attacks against RL mainly focus on either observation poisoning attacks or environment poisoning attacks. In this paper, we introduce a new class of attacks named action poisoning attacks, where an adversary can change the action signal selected by the agent. Compared with existing attack models, the attacker's ability in the proposed action poisoning attack model is more restricted, which brings some design challenges. We study the action poisoning attack in both white-box and black-box settings. We introduce an adaptive attack scheme called LCB-H, which works for most RL agents in the black-box setting. We prove that the LCB-H attack can force any efficient RL agent, whose dynamic regret scales sublinearly with the total number of steps taken, to choose actions according to a policy selected by the attacker very frequently, with only sublinear cost. In addition, we apply LCB-H attack against a popular model-free RL algorithm: UCB-H. We show that, even in the black-box setting, by spending only logarithm cost, the proposed LCB-H attack scheme can force the UCB-H agent to choose actions according to the policy selected by the attacker very frequently.

Due to the broad range of applications of stochastic multi-armed bandit model, understanding the effects of adversarial attacks and designing bandit algorithms robust to attacks are essential for the safe applications of this model. In this paper, we introduce a new class of attack named action-manipulation attack. In this attack, an adversary can change the action signal selected by the user. We show that without knowledge of mean rewards of arms, our proposed attack can manipulate Upper Confidence Bound (UCB) algorithm, a widely used bandit algorithm, into pulling a target arm very frequently by spending only logarithmic cost. To defend against this class of attacks, we introduce a novel algorithm that is robust to action-manipulation attacks when an upper bound for the total attack cost is given. We prove that our algorithm has a pseudo-regret upper bounded by $\mathcal{O}(\max\{\log T,A\})$, where $T$ is the total number of rounds and $A$ is the upper bound of the total attack cost., Comment: 13 pages, 7 figures, submitted to IEEE Transaction on Signal Processing