STEP: Success-Rate-Aware Trajectory-Efficient Policy Optimization (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing GRPO-based methods allocate sampling uniformly across tasks regardless of difficulty, propagate misleading learning signals and incur high sample-collection costs. |
| Approach: | They propose a framework that allocates sampling based on per-task success rates and performs fine-grained step-level optimization. |
| Outcome: | The proposed method improves sample efficiency and training stability over existing GRPO variants and three ablation variants on OSWorld and AndroidWorld. |
Similar Papers
TSPO: Breaking the Double Homogenization Dilemma in Multi-turn Search Policy Optimization (2026.findings-acl)
Copied to clipboard
Shichao Ma, Zhiyuan Ma, Ming Yang, Xiaofan Li, Xing Wu, Jintao Du, Yu Cheng, Weiqiang Wang, Qiliang Liu, Zhengyang Zhou, Yang Wang
| Challenge: | Large Language Models (LLMs) can solve complex tasks through iterative information retrieval. |
| Approach: | They propose a turn-level stage-aware policy optimization approach to solve this problem . they introduce a first-occurrence latent reward mechanism to allocate partial rewards . |
| Outcome: | Experiments show that TSPO outperforms state-of-the-art models on Qwen2.5-3B and 7B models. |
TA-GRPO-d: Trajectory-Aware GRPO for Optimizing Denoising Trajectories in Diffusion LLMs (2026.acl-long)
Copied to clipboard
| Challenge: | Existing dLLMs rely on fixed denoising schedules and cannot learn efficient unmasking orders. |
| Approach: | They propose a framework that transforms dLLM decoding into a trajectory-aware policy . it uses a confidence-gated denoising strategy that decides which tokens to unmask . |
| Outcome: | The proposed model can learn which tokens to unmask and how many to unmak per step . it can learn the output quality and efficiency of the decoding path itself . |
Experience-driven Multi-turn Reinforcement Learning for GUI Agents (2026.acl-long)
Copied to clipboard
Zhengxi Lu, Jiabo Ye, Fei Tang, Yongliang Shen, Haiyang Xu, Ziwei Zheng, Weiming Lu, Ming Yan, Fei Huang, Jun Xiao, Yueting Zhuang
| Challenge: | GUI agents have demonstrated remarkable progress in automating complex user interface interactions . training such agents for long-horizon tasks remains challenging due to limited rewards and prohibitive costs. |
| Approach: | They propose a method that leverages expert trajectories as environment experiences for on-policy multi-turn training. |
| Outcome: | The proposed method achieves significant gains over the base model with 1K public trajectories as RL experiences . it achieves competitive performance against strong baselines such as UI-TARS-7B and GPT-4o . |
Full-Step-DPO: Self-Supervised Preference Optimization with Step-wise Rewards for Mathematical Reasoning (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to improve long-chain mathematical reasoning focus on the first erroneous step, but ignore all other steps and rely heavily on external signals. |
| Approach: | They propose a DPO framework that leverages step-wise rewards from the entire reasoning chain instead of optimizing only the first erroneous step. |
| Outcome: | The proposed framework improves on in-domain and out-of-domain mathematical reasoning benchmarks. |
PVPO: Pre-Estimated Value-Based Policy Optimization for Agentic Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | extending grouping-based methods to agentic reasoning presents unique challenges . frequent environment interactions and tool invocations render intra-group advantage estimation unstable . |
| Approach: | They propose a grouping-based method that uses a single round of rollouts to stabilize advantage estimation. |
| Outcome: | a new RL framework outperforms grouping-based methods in retrieval tasks and advanced mathematical reasoning benchmarks. |
Step-GRPO: Internalizing Dynamic Early Exit for Efficient Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Large reasoning models that use long chain-of-thought excel at problem-solving but waste computational resources. |
| Approach: | They propose a framework that internalizes dynamic early-exit capabilities directly into the model. |
| Outcome: | The proposed framework reduces token consumption by 32.0% on a Qwen3-8B model compared to the vanilla model . |
DPEPO: Diverse Parallel Exploration Policy Optimization for LLM-based Agents (2026.acl-long)
Copied to clipboard
JunShuo Zhang, Chengrui Huang, Feng Guo, Zihan Li, Ke Shi, Menghua Jiang, Jiguo Yu, Shuo Shang, Shen Gao
| Challenge: | Existing approaches to large language model (LLM) agents that follow the sequential "reason-then-act" paradigm suffer from limited exploration and incomplete environmental understanding as they interact with only a single environment per step. |
| Approach: | They propose a paradigm that enables an agent to interact with multiple environments simultaneously and share cross-trajectory experiences. |
| Outcome: | The proposed paradigm achieves state-of-the-art (SOTA) success rates while maintaining comparable efficiency to strong sequential baselines. |
Think Outside the Policy: In-Context Steered Policy Optimization (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing Reinforcement Learning from Verifiable Rewards (RLVR) methods exhibit limited exploration due to reliance on on-policy rollouts which are limited to the current policy’s distribution, resulting in narrow trajectory diversity. |
| Approach: | They propose a framework that leverages the in-context learning capability of Large Reasoning Models to provide expert guidance using existing datasets. |
| Outcome: | The proposed framework improves RLVR performance and training stability on mathematical reasoning benchmarks. |
Empowering Multi-Turn Tool-Integrated Agentic Reasoning with Group Turn Policy Optimization (2026.acl-long)
Copied to clipboard
Yifeng Ding, Hung Le, Songyang Han, Kangrui Ruan, Zhenghui Jin, Varun Kumar, Zijian Wang, Anoop Deoras
| Challenge: | Current reinforcement learning methods suffer from coarse-grained, trajectory-level rewards that provide insufficient learning signals for complex multi-turn interactions, leading to training stagnation. |
| Approach: | They propose a novel RL algorithm for training large language models for multi-turn tool-integrated reasoning (TIR) that incorporates three innovations: turn-level reward assignment that provides fine-grained feedback for individual turns, return-based advantage estimation where normalized discounted returns are calculated as advantages, and self-supervised reward shaping that exploits self-supervision signals from generated code to densify sparse binary outcome-based rewards. |
| Outcome: | The proposed algorithm outperforms GRPO by 3.0% across diverse math reasoning benchmarks and improves grepo by 3.9% on commonsense reasoning and program synthesis tasks. |
StepSearch: Igniting LLMs Search Ability via Step-Wise Proximal Policy Optimization (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent work has demonstrated unprecedented capabilities in sophisticated linguistic comprehension and generative tasks. |
| Approach: | They propose a framework for search LLMs that trains with step-wise proximal policy optimization method to improve QA performance. |
| Outcome: | The proposed framework outperforms global-reward benchmarks on multi-hop QA with a stepwise proximal policy optimization method and richer and more detailed intermediate search rewards and token-level process supervision. |