Improving the Language Understanding Capabilities of Large Language Models Using Reinforcement Learning

• URL: http://arxiv.org/abs/2410.11020v4
• Date: Tue, 03 Jun 2025 13:16:21 GMT
• Title: Improving the Language Understanding Capabilities of Large Language Models Using Reinforcement Learning
• Authors: Bokai Hu, Sai Ashish Somayajula, Xin Pan, Pengtao Xie,
• Abstract summary: Proximal Policy Optimization (PPO) is a framework to improve the capabilities of large language models (LLMs)<n>PPO consistently outperforms supervised fine-tuning, yielding an average improvement of 6.3 points on GLUE.<n>This work highlights a promising direction for adapting LLMs to new tasks by reframing them as reinforcement learning problems.
• Score: 20.13007387453759
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Instruction-fine-tuned large language models (LLMs) under 14B parameters continue to underperform on natural language understanding (NLU) tasks, often trailing smaller models like BERT-base on benchmarks such as GLUE and SuperGLUE. Motivated by the success of reinforcement learning in reasoning tasks (e.g., DeepSeek), we explore Proximal Policy Optimization (PPO) as a framework to improve the NLU capabilities of LLMs. We frame NLU as a reinforcement learning environment, treating token generation as a sequence of actions and optimizing for reward signals based on alignment with ground-truth labels. PPO consistently outperforms supervised fine-tuning, yielding an average improvement of 6.3 points on GLUE, and surpasses zero-shot and few-shot prompting by 38.7 and 26.1 points, respectively. Notably, PPO-tuned models outperform GPT-4o by over 4\% on average across sentiment and natural language inference tasks, including gains of 7.3\% on the Mental Health dataset and 10.9\% on SIGA-nli. This work highlights a promising direction for adapting LLMs to new tasks by reframing them as reinforcement learning problems, enabling learning through simple end-task rewards rather than extensive data curation.

Related papers

• GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning [106.98018881499362]
  We introduce GEPA (Genetic-Pareto), a prompt that thoroughly incorporates natural language to learn high-level rules from trial and error.<n>GEPA samples system-level trajectories (e.g., reasoning, tool calls, and tool outputs) and reflects on them in natural language to diagnose problems.<n>It can often turn even just a few rollouts into a large quality gain.
  arXiv  Detail & Related papers  (2025-07-25T17:42:32Z)
• Critique-GRPO: Advancing LLM Reasoning with Natural Language and Numerical Feedback [59.078756231841574]
  Critique-GRPO is an online RL framework that integrates both natural language and numerical feedback for effective policy optimization.<n>We show Critique-GRPO consistently outperforms supervised learning and RL-based fine-tuning methods across eight challenging mathematical, STEM, and general reasoning tasks.
  arXiv  Detail & Related papers  (2025-06-03T17:39:02Z)
• Catastrophic Forgetting in LLMs: A Comparative Analysis Across Language Tasks [0.0]
  Large Language Models (LLMs) have significantly advanced Natural Language Processing (NLP)<n>This study evaluates the continual fine-tuning of various open-source LLMs on key NLU tasks.<n>Our results indicate that models such as Phi-3.5-mini exhibit minimal forgetting while maintaining strong learning capabilities.
  arXiv  Detail & Related papers  (2025-04-01T23:06:55Z)
• Adaptive Pruning for Large Language Models with Structural Importance Awareness [66.2690963378878]
  Large language models (LLMs) have significantly improved language understanding and generation capabilities. LLMs are difficult to deploy on resource-constrained edge devices due to their high computational and storage resource demands. We propose structurally-aware adaptive pruning (SAAP) to significantly reduce the computational and memory costs while maintaining model performance.
  arXiv  Detail & Related papers  (2024-12-19T18:08:04Z)
• GIVE: Structured Reasoning of Large Language Models with Knowledge Graph Inspired Veracity Extrapolation [108.2008975785364]
  Graph Inspired Veracity Extrapolation (GIVE) is a novel reasoning method that merges parametric and non-parametric memories to improve accurate reasoning with minimal external input.<n>GIVE guides the LLM agent to select the most pertinent expert data (observe), engage in query-specific divergent thinking (reflect), and then synthesize this information to produce the final output (speak)
  arXiv  Detail & Related papers  (2024-10-11T03:05:06Z)
• Large Language Models as Code Executors: An Exploratory Study [29.545321608864295]
  This paper pioneers the exploration of Large Language Models (LLMs) as code executors. We are the first to examine this feasibility across various LLMs, including OpenAI's o1, GPT-4o, GPT-3.5, DeepSeek, and Qwen-Coder. We introduce an Iterative Instruction Prompting (IIP) technique that processes code snippets line by line, enhancing the accuracy of weaker models by an average of 7.22%.
  arXiv  Detail & Related papers  (2024-10-09T08:23:22Z)
• VinePPO: Unlocking RL Potential For LLM Reasoning Through Refined Credit Assignment [66.80143024475635]
  We propose VinePPO, a straightforward approach to compute unbiased Monte Carlo-based estimates. We show that VinePPO consistently outperforms PPO and other RL-free baselines across MATH and GSM8K datasets.
  arXiv  Detail & Related papers  (2024-10-02T15:49:30Z)
• Minor SFT loss for LLM fine-tune to increase performance and reduce model deviation [9.506166330956082]
  We propose a training metric for SFT to measure the discrepancy between the optimized model and the original model, and a loss function MinorSFT that can increase the training effectiveness. In this article with insight from DPO and MinorDPO, we propose a training metric for SFT to measure the discrepancy between the optimized model and the original model, and a loss function MinorSFT that can increase the training effectiveness.
  arXiv  Detail & Related papers  (2024-08-20T08:32:44Z)
• Enhancing Reinforcement Learning with Label-Sensitive Reward for Natural Language Understanding [11.470005425117371]
  We propose a novel Reinforcement Learning framework enhanced with Label-sensitive Reward (RLLR) Our method aims to adeptly capture nuanced label-sensitive semantic features during RL, thereby enhancing natural language understanding. Experiments conducted on five diverse foundation models across eight tasks showcase promising results.
  arXiv  Detail & Related papers  (2024-05-30T07:19:31Z)
• Self-Play Preference Optimization for Language Model Alignment [75.83359213697854]
  Recent advancements suggest that directly working with preference probabilities can yield a more accurate reflection of human preferences. We propose a self-play-based method for language model alignment, which treats the problem as a constant-sum two-player game. Our approach, dubbed Self-Play Preference Optimization (SPPO), utilizes iterative policy updates to provably approximate the Nash equilibrium.
  arXiv  Detail & Related papers  (2024-05-01T17:59:20Z)
• Token-level Direct Preference Optimization [8.249403373337024]
  Fine-tuning pre-trained Large Language Models is essential to align them with human values and intentions. We introduce Token-level Direct Preference Optimization (TDPO), a novel approach to align LLMs with human preferences by optimizing policy at the token level.
  arXiv  Detail & Related papers  (2024-04-18T08:49:38Z)
• Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models [52.98743860365194]
  We propose a new fine-tuning method called Self-Play fIne-tuNing (SPIN) At the heart of SPIN lies a self-play mechanism, where the LLM refines its capability by playing against instances of itself. This sheds light on the promise of self-play, enabling the achievement of human-level performance in LLMs without the need for expert opponents.
  arXiv  Detail & Related papers  (2024-01-02T18:53:13Z)
• Using Natural Language Explanations to Improve Robustness of In-context Learning [35.18010811754959]
  Large language models (LLMs) can excel in many tasks via in-context learning (ICL) We investigate whether augmenting ICL with natural language explanations (NLEs) improves the robustness of LLMs on adversarial datasets.
  arXiv  Detail & Related papers  (2023-11-13T18:49:13Z)
• L3 Ensembles: Lifelong Learning Approach for Ensemble of Foundational Language Models [15.726224465017596]
  We propose an approach that focuses on extracting meaningful representations from unseen data and constructing a structured knowledge base. We conducted experiments on various NLP tasks to validate its effectiveness, including benchmarks like GLUE and SuperGLUE. The proposed L3 ensemble method increases the model accuracy by 4% 36% compared to the fine-tuned FLM.
  arXiv  Detail & Related papers  (2023-11-11T06:59:50Z)
• Black-Box Prompt Optimization: Aligning Large Language Models without Model Training [95.73262836039231]
  Large language models (LLMs) have shown impressive success in various applications. LLMs are often not well aligned with human intents, which calls for additional treatments on them. In this work, we take a different perspective -- Black-Box Prompt Optimization (BPO) -- to perform alignments.
  arXiv  Detail & Related papers  (2023-11-07T17:31:50Z)
• Improving Translation Faithfulness of Large Language Models via Augmenting Instructions [89.76691340615848]
  We propose SWIE (Segment-Weighted Instruction Embedding) and an instruction-following dataset OVERMISS. SWIE improves the model instruction understanding by adding a global instruction representation on the following input and response representations. OVERMISS improves model faithfulness by comparing over-translation and miss-translation results with the correct translation.
  arXiv  Detail & Related papers  (2023-08-24T09:32:29Z)
• Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting [65.00288634420812]
  Pairwise Ranking Prompting (PRP) is a technique to significantly reduce the burden on Large Language Models (LLMs) Our results are the first in the literature to achieve state-of-the-art ranking performance on standard benchmarks using moderate-sized open-sourced LLMs.
  arXiv  Detail & Related papers  (2023-06-30T11:32:25Z)
• LLM-Pruner: On the Structural Pruning of Large Language Models [65.02607075556742]
  Large language models (LLMs) have shown remarkable capabilities in language understanding and generation. We tackle the compression of LLMs within the bound of two constraints: being task-agnostic and minimizing the reliance on the original training dataset. Our method, named LLM-Pruner, adopts structural pruning that selectively removes non-critical coupled structures.
  arXiv  Detail & Related papers  (2023-05-19T12:10:53Z)
• Generating Training Data with Language Models: Towards Zero-Shot Language Understanding [35.92571138322246]
  Pretrained language models (PLMs) have demonstrated remarkable performance in various natural language processing tasks. We present a simple approach that uses both types of PLMs for fully zero-shot learning of NLU tasks. Our approach demonstrates strong performance across seven classification tasks of the GLUE benchmark.
  arXiv  Detail & Related papers  (2022-02-09T16:02:18Z)

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.