论文检索

Deep Learning · Large Language Models

Qingnan Ren, Zhen Fang, Shiting Huang, Yu Zeng, Lin Chen, Zehui Chen, Feng Zhao

Reward models are central to Reinforcement Learning from Human Feedback (RLHF), especially for open-ended tasks where evaluation is inherently multi-dimensional. Recent Generative Reward Models (GRMs) improve interpretability by producing natural-language rationales and task-specific evaluation principles. However, most existing GRMs generate principles after reading the actor's response, i.e., $Q+R \rightarrow P$. We show that this coupling induces Principle Drift: when the actor performs reward hacking (e.g., verbosity, self-aggrandizement, or hallucinated self-justifications), the reward model may shift its criteria to rationalize the response, yielding inflated scores that in turn reinforce hacking during RL. We propose IP-GRM (Independent Principle GRM), a two-stage framework that first generates principles solely from the question ($Q \rightarrow P$) and then evaluates the response conditioned on $(Q, R, P)$. This decoupling keeps criteria invariant to response content, producing more objective and stable reward signals. For efficient training, we further introduce a Principle Cache strategy that reuses principles within a group, improving GRPO throughput by 23.66\% while maintaining strict intra-group consistency. In GRPO training on creative writing, IP-GRM suppresses reward hacking and improves WritingBench and CreativeWriting-v3 by up to +4.6 and +7.1 points based on Qwen3-8B, achieving state-of-the-art performance among open-source models.

Deep Learning · Large Language Models

Penghui Qi, Xiangxin Zhou, Zichen Liu, Tianyu Pang, Chao Du, Min Lin, Wee Sun Lee

Reinforcement learning (RL) has become a cornerstone for fine-tuning Large Language Models (LLMs), with Proximal Policy Optimization (PPO) serving as the de facto standard algorithm. Despite its ubiquity, we argue that the core ratio clipping mechanism in PPO is structurally ill-suited for the large vocabularies inherent to LLMs. PPO constrains policy updates based on the probability ratio of sampled tokens, which serves as a noisy single-sample Monte Carlo estimate of the true policy divergence. This creates a sub-optimal learning dynamic: updates to low-probability tokens are aggressively over-penalized, while potentially catastrophic shifts in high-probability tokens are under-constrained, leading to training inefficiency and instability. To address this, we propose Divergence Proximal Policy Optimization (DPPO), which substitutes heuristic clipping with a more principled constraint based on a direct estimate of policy divergence (e.g., Total Variation or KL). To avoid huge memory footprint, we introduce the efficient Binary and Top-K approximations to capture the essential divergence with negligible overhead. Extensive empirical evaluations demonstrate that DPPO achieves superior stability and efficiency compared to existing methods, offering a more robust foundation for RL-based LLM fine-tuning.

Applications · Computer Vision

Zhenhua Ning, Xin Li, Jun Yu, Guangming Lu, Yaowei Wang, Wenjie Pei

While 3D Gaussian Splatting (3DGS) has demonstrated impressive real-time rendering performance, its efficacy remains constrained by a reliance on heuristic density control. Despite numerous refinements to these handcrafted rules, such methods inherently lack the flexibility to adapt to diverse scenes with complex geometries. In this paper, we propose a paradigm shift for density control from rigid heuristics to fully learnable policies. Specifically, we introduce LeGS, a framework that reformulates density control as a parameterized policy network optimized via Reinforcement Learning (RL). Central to our approach is the tailored effective reward function grounded in sensitivity analysis, which precisely quantifies the marginal contribution of individual Gaussians to reconstruction quality. To maintain computational tractability, we derive a closed-form solution that reduces the complexity of reward calculation from $O(N^2)$ to $O(N)$. Extensive experiments on the Mip-NeRF 360, Tanks \& Temples, and Deep Blending datasets demonstrate that LeGS significantly outperforms state-of-the-art methods, striking a superior balance between reconstruction quality and efficiency.

Reinforcement Learning · Deep RL

Zekang Wang, Zhe He, Borong Zhang, Edan Toledo, Steven Morad

How should we analyze memory in deep RL? We introduce tools for analyzing policies under partial observability and revealing how agents use memory to make decisions. To utilize these tools, we present POPGym Arcade, a collection of Atari-inspired, hardware-accelerated environments sharing a single observation and action space. Each environment provides fully and partially observable variants, enabling counterfactual studies on observability. We find that controlled studies are necessary for fair comparisons and identify a pathology where value functions smear credit over irrelevant history. Using this pathology, we demonstrate how out-of-distribution scenarios can contaminate memory, perturbing the policy far into the future.

Reinforcement Learning · Batch/Offline

Mingxuan Li, Junzhe Zhang, Elias Bareinboim

Expressive policies based on flow-matching have been successfully applied in reinforcement learning (RL) more recently due to their ability to model complex action distributions from offline data. These algorithms build on standard policy gradients, which assume that there is no unmeasured confounding in the data. However, this condition does not necessarily hold for pixel-based demonstrations when a mismatch exists between the demonstrator's and the learner's sensory capabilities, leading to implicit confounding biases in offline data. We address the challenge by investigating the problem of confounded observations in offline RL from a causal perspective. We develop a novel causal offline RL objective that optimizes policies' worst-case performance that may arise due to confounding biases. Based on this new objective, we introduce a practical implementation that learns expressive flow-matching policies from confounded demonstrations, employing a deep discriminator to assess the discrepancy between the target policy and the nominal behavioral policy. Experiments across 25 pixel-based tasks demonstrate that our proposed confounding-robust augmentation procedure achieves a success rate 120\% that of confounding-unaware, state-of-the-art offline RL methods.

Reinforcement Learning · Deep RL

Lirui Luo, Guoxi Zhang, Hongming Xu, Cong Fang, Qing Li

In DRL, an agent is trained from a stream of experience. In a continual learning setting, such agents can suffer from \emph{plasticity loss}: their ability to learn new skills from new experiences diminishes over training. Recently, Mixture-of-Experts (MoE) networks have been reported to enable scaling laws and facilitate the learning of diverse skills. However, in continual reinforcement learning settings, their performance can degenerate as learning proceeds, indicating a loss of plasticity. To address this, building on Neural Tangent Kernel (NTK) theory, we formalize the plasticity loss in MoE policies as a loss of \emph{spectral plasticity}. We then derive a tractable proxy for spectral plasticity, one expressible in terms of individual expert feature matrices. Leveraging this proxy, we introduce \emph{SPHERE}, a practical Parseval penalty tailored for MoE-based policies that alleviates the loss of spectral plasticity. On MetaWorld and HumanoidBench, SPHERE improves average success under continual RL by 133\% and 50\% over an unregularized MoE baseline, while maintaining higher spectral plasticity throughout training.

Deep Learning · Large Language Models

Yun Qu, Qi Wang, Yixiu Mao, Heming Zou, Yuhang Jiang, Weijie Liu, Clive Bai, Kai Yang, Yangkun Chen, Saiyong Yang 等

Reinforcement learning enhances the reasoning capabilities of large language models but often involves high computational costs due to rollout-intensive optimization. Online prompt selection presents a plausible solution by prioritizing informative prompts to improve training efficiency. However, current methods either depend on costly, exact evaluations or construct prompt-specific predictive models lacking generalization across prompts. This study introduces Generalizable Predictive Prompt Selection (GPS), which performs Bayesian inference towards prompt difficulty using a lightweight generative model trained on the shared optimization history. Intermediate-difficulty prioritization and history-anchored diversity are incorporated into the batch acquisition principle to select informative prompt batches. The small predictive model also generalizes at test-time for efficient computational allocation. Experiments across varied reasoning benchmarks indicate GPS's substantial improvements in training efficiency, final performance, and test-time efficiency over superior baseline methods. The code is available at https://anonymous.4open.science/r/GPS-ICML.

Deep Learning · Large Language Models

Zishang Jiang, tingyun li, Jinyi Han, Xinyi Wang, Mengyun Qiao, Yizhou Ying, Xiaojun Meng, Jiansheng Wei, Jiaqing Liang, Yanghua Xiao

Reinforcement learning (RL) has become a widely adopted technique for improving large language models (LLMs) on complex tasks. Despite this progress, existing RL methods still face challenges in training agents with longer-horizon interactions. One major bottleneck is distinguishing the contribution of different actions in long-horizon interaction, leading to high optimization variance. To address this, we introduce a novel policy gradient method, Hindsight Policy Optimization (HPO), that projects both the current policy distribution and the hindsight distribution into an intent space and extracts low-variance learning signals from the Wasserstein distance between them. We theoretically and empirically show that aggregating semantically similar states and actions in the intent space yields a bounded-variance estimator and improves policy performance stably. Our code is available online.

Deep Learning · Large Language Models

Zixuan Huang, Xin Xia, Yuxi Ren, Jianbin Zheng, Xuefeng Xiao, Hongyan Xie, Huaqiu Li, Songshi Liang, Zhongxiang Dai, Fuzhen Zhuang 等

Reinforcement Learning from Human Feedback (RLHF) is a pivotal technique for aligning large language models (LLMs) with human preferences, yet it is susceptible to reward overoptimization, in which policy models overfit to the reward model, exploit spurious reward patterns instead of faithfully capturing human intent. Prior mitigations primarily relies on surface semantic information and fails to efficiently address the misalignment between the reward model (RM) and the policy model caused by continuous policy distribution shifts. This inevitably leads to an increasing reward discrepancy, exacerbating reward overoptimization. To address these limitations, we introduce R2M (Real-Time Aligned Reward Model), a novel lightweight RLHF framework. R2M goes beyond vanilla reward models that solely depend on the semantic representations of a pretrained LLM. Instead, it leverages the evolving hidden states of the policy (namely policy feedback) to align with the real-time distribution shift of the policy during the RL process. This work points to a promising new direction for improving the performance of reward models through real-time utilization of feedback from policy models.

Deep Learning · Large Language Models

Jian Mu, Tianyi Lin, Chengwei Qin, Zhongxiang Dai, Yao Shu

Large language models are increasingly deployed in multi-turn interactive settings where users or environments can iteratively provide lightweight feedback. Unfortunately, optimizing such behavior presents a sharp dilemma in practice: online reinforcement learning is able to effectively address multi-turn dynamics but is prohibitively expensive due to the cost of generating full correction trajectories at every update, whereas offline supervised fine-tuning (SFT) is efficient but suffers from distribution shift and behavioral collapse. To this end, we novelly propose DRIFT (Decoupled Rollouts and Importance-Weighted Fine-Tuning), a framework that operationalizes the theoretical insight that the KL-regularized RL objective is mathematically equivalent to importance-weighted supervised learning. DRIFT decouples rollout from optimization by sampling offline interaction trajectories from a fixed reference policy, deriving return-based importance weights, and optimizing the policy via weighted SFT on the resulting dataset. Empirically, we demonstrate that DRIFT matches or exceeds the performance of multi-turn reinforcement learning baselines while maintaining the training efficiency and simplicity of standard supervised fine-tuning.

Deep Learning · Foundation Models

Tejas Krishnan, Sumeet Motwani, Charles London, Suhaas Bhat, Huitian Jiao, Phil Torr, Riashat Islam, Christopher Summerfield, Christian Schroeder de Witt, Qilong Gu 等

Reinforcement learning with verifiable rewards (RLVR) on foundation models has led to significant improvements in math and code generation. Extending these gains to open-ended domains remains challenging: ground-truth verification is unavailable, human annotation is expensive, and learnt reward models are prone to reward hacking. We introduce Rubric Curriculum RL (RcRL), a self-improvement method for creative short-fiction writing that requires no new data, human annotations, or stronger teacher models. RcRL exploits the generation-verification gap: it is easier to judge whether work is creative than to produce something creative. While this gap exists across open-ended domains, exploiting it for RL is challenging due to reward hacking. During training, we use pairwise preferences against a curriculum of rubric criteria, which provides a more stable signal than absolute scoring while reducing reward hacking against a stationary objective. Unlike baseline methods, which plateau or collapse within a few dozen steps, our approach preserves output entropy and shows improvements over 1000+ training steps. In human evaluations, RcRL-trained models achieve a 70.5% win rate and demonstrate consistent gains across multiple creative writing benchmarks and judges.

Deep Learning · Large Language Models

Pouya Behzadifar, Parshin Shojaee, Sanchit Kabra, Kazem Meidani, Chandan Reddy

Finding mathematical relations underlying natural phenomena is a fundamental task in scientific discovery. Recent advances in evolutionary search with Large Language Models (LLMs) show great promise by leveraging their embedded scientific knowledge. However, discovering governing equations remains challenging due to vast combinatorial hypothesis spaces with exponentially many possible relations. Existing LLM-based approaches treat LLMs as static hypothesis generators unaware of the observed scientific system, leading to suboptimal and inefficient exploration that over-relies on internal priors. To address this, we introduce \emph{Decompose, Adapt, and Evolve} (\textbf{DecAEvolve}), a framework that combines granular feedback from symbolic term decomposition with LLM refinement through reinforcement learning fine-tuning. DecAEvolve unifies symbolic decomposition with test-time RL adaptation, enabling adaptive rather than static hypothesis generation. Our experiments across diverse scientific benchmarks demonstrate that DecAEvolve significantly improves both the accuracy of discovered equations and the efficiency of the discovery process, reducing error by up to an order of magnitude compared to state-of-the-art baselines.

Deep Learning · Large Language Models

Zanlin Ni, Shenzhi Wang, Yang Yue, Tianyu Yu, Weilin Zhao, Yeguo Hua, Tianyi Chen, Jun Song, YuCheng, Bo Zheng 等

Diffusion Large Language Models (dLLMs) break the rigid left-to-right constraint of traditional LLMs, enabling token generation in arbitrary orders. Intuitively, this flexibility implies a solution space that strictly supersets the fixed autoregressive trajectory, theoretically unlocking superior reasoning potential. Indeed, for specific constraint satisfaction tasks (e.g., sudoku puzzles), this capability has proven to be highly advantageous. However, in this paper, we reveal that for general reasoning tasks (e.g., mathematics and coding), arbitrary order generation may in fact limit the reasoning potential of dLLMs. We find that dLLMs tend to exploit this order flexibility to bypass high-uncertainty tokens that are crucial for exploration, leading to a premature collapse of solution coverage. This observation motivates a rethink of RL approaches for dLLMs, where considerable complexities, such as handling combinatorial trajectories and intractable likelihoods, are often devoted to preserving this flexibility. We demonstrate that effective reasoning can be better elicited by simply forgoing arbitrary order and applying standard Group Relative Policy Optimization (GRPO) instead. Our approach, **JustGRPO**, is minimalist yet surprisingly effective (e.g., 89.1% accuracy on GSM8K) while fully retaining the parallel decoding ability of dLLMs.

Reinforcement Learning · Deep RL

Julien Brosseit, Jasper Hoffmann, Joschka Boedecker

Early exit neural networks, which adapt computation to input complexity, have proven effective in supervised learning but remain largely unexplored in deep reinforcement learning (DRL). In this paper, we propose the use of Budgeted EXit Actor (BEXA), which is a novel actor-critic architecture that integrates early exit branches into the actor network. These branches are trained via the underlying DRL method and use a constrained value-based criterion to decide when to exit, allowing the policy to dynamically adjust its computation. BEXA is general, easy to tune and compatible with any off-policy actor-critic method. We evaluate BEXA using different DRL methods such as SAC and TD3 on a suite of MuJoCo tasks. Our results demonstrate a substantial improvement in inference efficiency with minimal or no loss in performance. These findings highlight early exits as a promising direction for improving computational efficiency in DRL.

Reinforcement Learning · Deep RL

Cheng Xin, Shuo He, Lang Feng, Haiyang Xu, Ming Yan, Lei Feng, Bo An

Group-based reinforcement learning (RL) methods have achieved remarkable success in improving the performance of large language models (LLMs) and have been rapidly extended to agentic tasks. However, their credit assignment relies heavily on coarse-grained trajectory-level attribution according to final outcomes, making it difficult to capture the contribution of individual steps, such as valuable steps obscured within failed trajectories. To uncover latent information and enable more faithful step-level credit assignment, we propose Graph-based Group Policy Optimization (GraphGPO), which first aggregates all rollout trajectories into a unified state-transition graph and then estimates the distance from each state to the task goal using the global information encoded in the graph. Finally, GraphGPO assigns credit to each edge by estimating a graph-based advantage, based on how much the transition reduces the distance to the task goal. In this way, GraphGPO significantly improves training efficiency and achieves state-of-the-art performance across a range of challenging benchmarks.

Deep Learning · Large Language Models

Avinandan Bose, Stella Li, Faeze Brahman, Pang Wei Koh, Simon Du, Yulia Tsvetkov, Maryam Fazel, Lin Xiao, Asli Celikyilmaz

Cold-start personalization requires inferring preferences from minimal interaction when no user-specific historical data is available. The space of possible preferences is vast, yet users care about only a sparse subset and rarely articulate them upfront; combined with limited interaction budgets, this makes preference elicitation challenging. Our key insight is that preferences exhibit predictable structure across populations; e.g., users who want detailed explanations often also value worked examples. We propose PEP (Preference Elicitation with Priors), a principled system decomposition framework for cold-start personalization: learning a structured world model of preference correlations offline using latent variables, then performing Bayesian inference online without retraining. Even simple belief model instantiations (e.g., linear regression) substantially outperform end-to-end RL. Across medical, mathematical, social, and commonsense reasoning, PEP achieves 80.8% alignment with ground-truth user preferences versus 68.5% for RL, requires 3-5× fewer interactions, and adapts twice as often. Our contribution is a principled decomposition of cold-start personalization that makes Bayesian preference elicitation practical at scale for LLM systems.

Deep Learning · Large Language Models

Xiaoling Zhou, Shuaiyu Zhou, Zhemg Lee, Tao Chen, Xirui Li, Peng Chen, Jie Jiang, Wei Ye, Shikun Zhang

Although reinforcement learning (RL) enhances the reasoning capabilities of large language models (LLMs), it is primarily learned from the model's self-generated distribution, limiting its ability to acquire reasoning skills beyond its initial knowledge. To overcome this, we propose a Difficulty-Aware Learning Strategy Allocation (DALSA) framework, which adaptively assigns appropriate learning strategies to samples based on their difficulty signals. DALSA is built on the key insight that samples beyond models' knowledge scope are better addressed through supervised fine-tuning (SFT), while those within the boundary but insufficiently mastered benefit more from RL, and well-learned samples are discarded to avoid redundant updates. To realize this principle, we extract a series of difficulty-aware training characteristics and employ a learnable strategy allocator to dynamically determine the optimal learning strategy for each sample based on its training dynamics. The allocator and the LLM are alternately optimized, enabling adaptive strategy allocation. Furthermore, two regularization techniques, anti-curriculum weighting and adversarial label smoothing, are integrated to alleviate the inherent limitations of RL and SFT, backed by comprehensive theoretical analyses. Extensive experiments on ten LLMs ranging from 1.5B to 70B across various tasks indicate that DALSA consistently outperforms baselines under both full and parameter-efficient fine-tuning settings.

Deep Learning · Large Language Models

Fanqi Kong, Jiayi Zhang, Mingyi Deng, Chenglin Wu, Yuyu Luo, Bang Liu

Real-world user requests to LLM agents are often underspecified. Agents must interact to acquire missing information and make correct downstream decisions. However, current multi-turn GRPO-based methods often rely on trajectory-level reward computation, which leads to credit assignment problems and insufficient advantage signals within rollout groups. A feasible approach is to identify valuable interaction turns at a fine granularity to drive more targeted learning. To address this, we introduce InfoPO, which frames multi-turn interaction as a process of active uncertainty reduction and computes an information-gain reward that credits turns whose feedback measurably changes the agent’s subsequent action distribution compared to a masked-feedback counterfactual. It then combines this signal with task outcomes via an adaptive variance-gated fusion to identify information importance while maintaining task oriented goal direction. Across diverse tasks including intent clarification, collaborative coding, and tool-augmented decision making, InfoPO consistently outperforms prompting and multi-turn RL baselines. It also demonstrates robustness under user simulator shifts and generalizes effectively to environment interactive tasks. Overall, InfoPO provides a principled and scalable mechanism for optimizing complex agent user collaboration.

Deep Learning · Large Language Models

Lucas Dionisopoulos, Nicklas Majamaki, Prithviraj Ammanabrolu

We study how reasoning evolves in a language model -- from supervised fine-tuning (SFT) to reinforcement learning (RL) -- by analyzing how a set of theoretically-inspired datasets impacts language model performance in chess. We find that fine-tuning a model to directly predict the best move leads to effective RL and the strongest downstream performance -- however, the RL stage elicits \textit{unfaithful} reasoning (reasoning inconsistent with the chosen move). Alternatively, training on multi-move trajectories yields comparable downstream performance with faithful reasoning and more stable RL. We show that RL induces a substantial positive shift in the distribution of move quality and reduces hallucination rates as a side effect. Finally, we find several SFT-checkpoint metrics -- metrics spanning evaluation performance, hallucination rates, and reasoning quality -- to be predictive of post-RL model performance. We release checkpoints and final models as well as training data, evaluations, and code that allowed us to surpass leading open-source reasoning models in chess with a 7B-parameter model.

Reinforcement Learning · Deep RL

Chen Jia

We propose to model text generation as a continuous-time latent dynamical process, where token generation is formulated as a Markov Decision Process whose internal state evolves via a neural ODE. This formulation bridges discrete token sequences and continuous semantic evolution, providing a theoretically grounded approach for coherent long-range generation. The framework is optimized via reinforcement learning, maximizing a composite objective that integrates task-specific rewards with knowledge distillation from a powerful pre-trained language model. Experiments demonstrate that our method, Continuous-Time Latent Language Model (CT-LLM), outperforms discrete baselines in generation coherence and long-context performance, offering a new paradigm for fluid and controllable language generation.