Data-Efficient RLVR via Off-Policy Influence Guidance

• URL: http://arxiv.org/abs/2510.26491v1
• Date: Thu, 30 Oct 2025 13:40:52 GMT
• Title: Data-Efficient RLVR via Off-Policy Influence Guidance
• Authors: Erle Zhu, Dazhi Jiang, Yuan Wang, Xujun Li, Jiale Cheng, Yuxian Gu, Yilin Niu, Aohan Zeng, Jie Tang, Minlie Huang, Hongning Wang,
• Abstract summary: This work proposes a theoretically-grounded approach using influence functions to estimate the contribution of each data point to the learning objective.<n>We develop textbfCurriculum textbfRL with textbfOff-textbfPolicy textInfluence guidance (textbfCROPI), a multi-stage RL framework that iteratively selects the most influential data for the current policy.
• Score: 84.60336960383867
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Data selection is a critical aspect of Reinforcement Learning with Verifiable Rewards (RLVR) for enhancing the reasoning capabilities of large language models (LLMs). Current data selection methods are largely heuristic-based, lacking theoretical guarantees and generalizability. This work proposes a theoretically-grounded approach using influence functions to estimate the contribution of each data point to the learning objective. To overcome the prohibitive computational cost of policy rollouts required for online influence estimation, we introduce an off-policy influence estimation method that efficiently approximates data influence using pre-collected offline trajectories. Furthermore, to manage the high-dimensional gradients of LLMs, we employ sparse random projection to reduce dimensionality and improve storage and computation efficiency. Leveraging these techniques, we develop \textbf{C}urriculum \textbf{R}L with \textbf{O}ff-\textbf{P}olicy \text{I}nfluence guidance (\textbf{CROPI}), a multi-stage RL framework that iteratively selects the most influential data for the current policy. Experiments on models up to 7B parameters demonstrate that CROPI significantly accelerates training. On a 1.5B model, it achieves a 2.66x step-level acceleration while using only 10\% of the data per stage compared to full-dataset training. Our results highlight the substantial potential of influence-based data selection for efficient RLVR.

Related papers

• Towards High Data Efficiency in Reinforcement Learning with Verifiable Reward [54.708851958671794]
  We propose a Data-Efficient Policy Optimization pipeline that combines optimized strategies for both offline and online data selection.<n>In offline phase, we curate a high-quality subset of training samples based on diversity, influence, and appropriate difficulty.<n>During online RLVR training, we introduce a sample-level explorability metric to dynamically filter samples with low exploration potential.
  arXiv  Detail & Related papers  (2025-09-01T10:04:20Z)
• LearnAlign: Reasoning Data Selection for Reinforcement Learning in Large Language Models Based on Improved Gradient Alignment [14.655048266761783]
  Reinforcement learning (RL) has become a key technique for enhancing LLMs' reasoning abilities, yet its data inefficiency remains a major bottleneck.<n>We present LearnAlign, which intelligently selects the learnable and representative training reasoning data for RL post-training.<n> Experiments across three mathematical reasoning benchmarks demonstrate that our method significantly reduces training data requirements.
  arXiv  Detail & Related papers  (2025-06-13T06:05:58Z)
• The Impact of On-Policy Parallelized Data Collection on Deep Reinforcement Learning Networks [12.82803159923457]
  parallel actors for data collection has been an effective technique used in reinforcement learning algorithms.<n>We conduct an empirical analysis of trade-offs on PPO, one of the most popular RL algorithms that uses parallel actors.<n>Our analyses indicate that larger dataset sizes can increase final performance across a variety of settings.
  arXiv  Detail & Related papers  (2025-06-03T21:27:17Z)
• ActiveDPO: Active Direct Preference Optimization for Sample-Efficient Alignment [94.36403843133616]
  Using human preferences to align large language models (LLMs) has significantly improved their performance in various downstream tasks.<n>Existing methods either lack a strong theoretical foundation or depend on restrictive reward function assumptions.<n>We propose an algorithm, ActiveDPO, that uses a theoretically grounded data selection criterion for non-linear reward functions.
  arXiv  Detail & Related papers  (2025-05-25T17:42:52Z)
• Enhancing Training Data Attribution with Representational Optimization [57.61977909113113]
  Training data attribution methods aim to measure how training data impacts a model's predictions.<n>We propose AirRep, a representation-based approach that closes this gap by learning task-specific and model-aligned representations explicitly for TDA.<n>AirRep introduces two key innovations: a trainable encoder tuned for attribution quality, and an attention-based pooling mechanism that enables accurate estimation of group-wise influence.
  arXiv  Detail & Related papers  (2025-05-24T05:17:53Z)
• UFO-RL: Uncertainty-Focused Optimization for Efficient Reinforcement Learning Data Selection [42.9272996371658]
  Single-pass uncertainty estimation is used to identify informative data instances, achieving up to 185x faster data evaluation.<n>Experiments show that training with just 10% of data selected by UFO-RL yields performance comparable to or surpassing full-data training.
  arXiv  Detail & Related papers  (2025-05-18T15:14:58Z)
• LEAD: Iterative Data Selection for Efficient LLM Instruction Tuning [22.242445543184264]
  We propose LEAD, an efficient iterative data selection framework that accurately estimates sample utility entirely within the standard training loop.<n>Experiments show that LEAD significantly outperforms state-of-the-art methods, improving average model performance by 6.1%-10.8% while using only 2.5% of the training data and reducing overall training time by 5-10x.
  arXiv  Detail & Related papers  (2025-05-12T10:57:51Z)
• ALinFiK: Learning to Approximate Linearized Future Influence Kernel for Scalable Third-Party LLM Data Valuation [11.36712576361739]
  Large Language Models (LLMs) heavily rely on high-quality training data, making data valuation crucial for optimizing model performance.<n>We introduce a linearized future influence kernel (LinFiK), which assesses the value of individual data samples.<n>We propose ALinFiK, a learning strategy to approximate LinFiK, enabling scalable data valuation.
  arXiv  Detail & Related papers  (2025-03-02T22:51:12Z)
• Value-Based Deep RL Scales Predictably [100.21834069400023]
  We show that value-based off-policy RL methods are predictable despite community lore regarding their pathological behavior.<n>We validate our approach using three algorithms: SAC, BRO, and PQL on DeepMind Control, OpenAI gym, and IsaacGym.
  arXiv  Detail & Related papers  (2025-02-06T18:59:47Z)
• Capturing the Temporal Dependence of Training Data Influence [100.91355498124527]
  We formalize the concept of trajectory-specific leave-one-out influence, which quantifies the impact of removing a data point during training.<n>We propose data value embedding, a novel technique enabling efficient approximation of trajectory-specific LOO.<n>As data value embedding captures training data ordering, it offers valuable insights into model training dynamics.
  arXiv  Detail & Related papers  (2024-12-12T18:28:55Z)
• Optimizing LLMs with Direct Preferences: A Data Efficiency Perspective [4.548047308860141]
  This study investigates the impact of different type of preference data on model performance. It aims to reduce their dependency on extensive amounts of preference data, which is expensive to collect.
  arXiv  Detail & Related papers  (2024-10-22T00:11:41Z)
• Latent-Variable Advantage-Weighted Policy Optimization for Offline RL [70.01851346635637]
  offline reinforcement learning methods hold the promise of learning policies from pre-collected datasets without the need to query the environment for new transitions. In practice, offline datasets are often heterogeneous, i.e., collected in a variety of scenarios. We propose to leverage latent-variable policies that can represent a broader class of policy distributions. Our method improves the average performance of the next best-performing offline reinforcement learning methods by 49% on heterogeneous datasets.
  arXiv  Detail & Related papers  (2022-03-16T21:17:03Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.