LIMR: Less is More for RL Scaling

• URL: http://arxiv.org/abs/2502.11886v1
• Date: Mon, 17 Feb 2025 15:13:29 GMT
• Title: LIMR: Less is More for RL Scaling
• Authors: Xuefeng Li, Haoyang Zou, Pengfei Liu,
• Abstract summary: We introduce Learning Impact Measurement (LIM), an automated method to evaluate and prioritize training samples. Our method achieves comparable or even superior performance using only 1,389 samples versus the full 8,523 samples dataset. For reproducible research and future innovation, we are open-sourcing LIMR, including implementation of LIM, training and evaluation code, curated datasets, and trained models.
• Score: 25.477841726836836
• License:
• Abstract: In this paper, we ask: what truly determines the effectiveness of RL training data for enhancing language models' reasoning capabilities? While recent advances like o1, Deepseek R1, and Kimi1.5 demonstrate RL's potential, the lack of transparency about training data requirements has hindered systematic progress. Starting directly from base models without distillation, we challenge the assumption that scaling up RL training data inherently improves performance. we demonstrate that a strategically selected subset of just 1,389 samples can outperform the full 8,523-sample dataset. We introduce Learning Impact Measurement (LIM), an automated method to evaluate and prioritize training samples based on their alignment with model learning trajectories, enabling efficient resource utilization and scalable implementation. Our method achieves comparable or even superior performance using only 1,389 samples versus the full 8,523 samples dataset. Notably, while recent data-efficient approaches (e.g., LIMO and s1) show promise with 32B-scale models, we find it significantly underperforms at 7B-scale through supervised fine-tuning (SFT). In contrast, our RL-based LIMR achieves 16.7% higher accuracy on AIME24 and outperforms LIMO and s1 by 13.0% and 22.2% on MATH500. These results fundamentally reshape our understanding of RL scaling in LLMs, demonstrating that precise sample selection, rather than data scale, may be the key to unlocking enhanced reasoning capabilities. For reproducible research and future innovation, we are open-sourcing LIMR, including implementation of LIM, training and evaluation code, curated datasets, and trained models at https://github.com/GAIR-NLP/LIMR.

Related papers

• S$^2$R: Teaching LLMs to Self-verify and Self-correct via Reinforcement Learning [51.84977135926156]
  We introduce S$2$R, an efficient framework that enhances LLM reasoning by teaching models to self-verify and self-correct during inference. Our results demonstrate that Qwen2.5-math-7B achieves an accuracy improvement from 51.0% to 81.6%, outperforming models trained on an equivalent amount of long-CoT distilled data.
  arXiv  Detail & Related papers  (2025-02-18T13:40:22Z)
• What Do Learning Dynamics Reveal About Generalization in LLM Reasoning? [83.83230167222852]
  We find that a model's generalization behavior can be effectively characterized by a training metric we call pre-memorization train accuracy. By connecting a model's learning behavior to its generalization, pre-memorization train accuracy can guide targeted improvements to training strategies.
  arXiv  Detail & Related papers  (2024-11-12T09:52:40Z)
• DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning [61.10299147201369]
  This paper introduces a novel autonomous RL approach, called DigiRL, for training in-the-wild device control agents. We build a scalable and parallelizable Android learning environment equipped with a VLM-based evaluator. We demonstrate the effectiveness of DigiRL using the Android-in-the-Wild dataset, where our 1.3B VLM trained with RL achieves a 49.5% absolute improvement.
  arXiv  Detail & Related papers  (2024-06-14T17:49:55Z)
• Getting More Juice Out of the SFT Data: Reward Learning from Human Demonstration Improves SFT for LLM Alignment [65.15914284008973]
  We propose to leverage an Inverse Reinforcement Learning (IRL) technique to simultaneously build an reward model and a policy model. We show that the proposed algorithms converge to the stationary solutions of the IRL problem. Our results indicate that it is beneficial to leverage reward learning throughout the entire alignment process.
  arXiv  Detail & Related papers  (2024-05-28T07:11:05Z)
• Monte Carlo Tree Search Boosts Reasoning via Iterative Preference Learning [55.96599486604344]
  We introduce an approach aimed at enhancing the reasoning capabilities of Large Language Models (LLMs) through an iterative preference learning process. We use Monte Carlo Tree Search (MCTS) to iteratively collect preference data, utilizing its look-ahead ability to break down instance-level rewards into more granular step-level signals. The proposed algorithm employs Direct Preference Optimization (DPO) to update the LLM policy using this newly generated step-level preference data.
  arXiv  Detail & Related papers  (2024-05-01T11:10:24Z)
• How to Train Data-Efficient LLMs [56.41105687693619]
  We study data-efficient approaches for pre-training language models (LLMs) We find that Ask-LLM and Density sampling are the best methods in their respective categories. In our comparison of 19 samplers, involving hundreds of evaluation tasks and pre-training runs, we find that Ask-LLM and Density are the best methods in their respective categories.
  arXiv  Detail & Related papers  (2024-02-15T02:27:57Z)
• Task Aware Modulation using Representation Learning: An Approach for Few Shot Learning in Environmental Systems [15.40286222692196]
  TAM-RL is a novel framework for few-shot learning in heterogeneous systems. We evaluate TAM-RL on two real-world environmental datasets.
  arXiv  Detail & Related papers  (2023-10-07T07:55:22Z)
• Complementary Ensemble Learning [1.90365714903665]
  We derive a technique to improve performance of state-of-the-art deep learning models. Specifically, we train auxiliary models which are able to complement state-of-the-art model uncertainty.
  arXiv  Detail & Related papers  (2021-11-09T03:23:05Z)
• PC-MLP: Model-based Reinforcement Learning with Policy Cover Guided Exploration [15.173628100049129]
  This work studies a model-based algorithm for both Kernelized Regulators (KNR) and linear Markov Decision Processes (MDPs) For both models, our algorithm guarantees sample complexity and only uses access to a planning oracle. Our method can also perform reward-free exploration efficiently.
  arXiv  Detail & Related papers  (2021-07-15T15:49:30Z)

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.