| Challenge: | Recent advances in large language models (LLMs) have catalyzed the development of autonomous agents capable of executing complex, multi-turn tasks. |
| Approach: | They propose a framework for agentic reinforcement learning that integrates turn-level tree search with tree search to address key challenges. |
| Outcome: | The proposed framework addresses key challenges: limited exploration diversity, sparse credit assignment, and misaligned policy optimization. |
Similar Papers
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. |
Turn-PPO: Turn-Level Advantage Estimation with PPO for Improved Multi-Turn RL in Agentic LLMs (2026.findings-eacl)
Copied to clipboard
| Challenge: | Reinforcement learning (RL) has re-emerged as a natural approach for training interactive LLM agents in real-world environments. |
| Approach: | They propose a variant that operates on a turn-level MDP formulation, instead of the commonly used token-level one. |
| Outcome: | The proposed method is more robust than the widely used GRPO algorithm and more efficient than token-level MDPs. |
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. |
SPARK: Strategic Policy-Aware Exploration via Dynamic Branching for Long-Horizon Agentic Learning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for training large language models waste computation budget on trivial steps while failing to guarantee sample quality. |
| Approach: | They propose a framework that selectively branches at critical decision states for resource-efficient exploration. |
| Outcome: | The proposed framework activates adaptive branching exploration at critical decision states to probe promising trajectories, thereby achieving precise resource allocation that prioritizes sampling quality over blind coverage. |
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. |
Proximity-Based Multi-Turn Optimization: Practical Credit Assignment for LLM Agent Training (2026.acl-industry)
Copied to clipboard
| Challenge: | Existing group-based policy optimization methods rely on statistical deviation within discrete batches, misallocating credit when task difficulty fluctuates. |
| Approach: | They propose a framework for multi-turn LLM agents that integrates global context . they propose GRPO, which integrates success-rate-aware modulation and proximity-based soft aggregation . |
| Outcome: | The proposed framework yields performance gains over existing baselines with negligible computational cost. |
KARL: Reinforcement Learning for LLM Agents on Multi-Turn Knowledge-Intensive Agentic Tasks (2026.acl-long)
Copied to clipboard
Xueqiao Sun, Xiao Liu, Bowen Lv, Hanchen Zhang, Bohao Jing, Zehan Qi, Yifan Xu, Yuxiao Dong, Jie Tang
| Challenge: | Large Language Models have shown remarkable potential as autonomous agents, but their effectiveness in knowledge-intensive tasks remains limited by passive knowledge utilization. |
| Approach: | They propose a framework that enables LLM agents to dynamically explore structured knowledge sources through multi-turn interactions. |
| Outcome: | The proposed framework outperforms existing retrieval-augmented approaches on knowledge graph and database tasks while maximizing tool-use behaviors end-to-end. |
Learning from the Irrecoverable: Error-Localized Policy Optimization for Tool-Integrated LLM Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Tool-integrated reasoning (TIR) enables LLM agents to solve tasks through planning, tool use, and iterative revision, but outcome-only reinforcement learning suffers from sparse, delayed rewards and weak step-level credit assignment. |
| Approach: | They propose a tool-integrated reasoning approach that localizes the first irrecoverable step and leverages it for fine-grained credit assignment. |
| Outcome: | The proposed algorithm outperforms strong Agentic RL benchmarks in math, science QA, and code execution with additional gains in Pass@K and Major@K scaling, rollout ranking quality, and tool-call efficiency. |
MGPO: Thinking with Images via Multi-Turn Grounding-Based Reinforcement Learning (2026.findings-acl)
Copied to clipboard
| Challenge: | State-of-the-art large multimodal models face challenges when processing high-resolution images, as these inputs are converted into enormous visual tokens, many of which are irrelevant to the downstream task. |
| Approach: | They propose a multi-turn grounding-based policy optimization framework that enables LMMs to iteratively focus on key visual regions by automatically cropping sub-images based on model-predicted grounding coordinates within a multiple-turn conversation framework. |
| Outcome: | The proposed framework improves on Qwen2.5-VL-7B with 21K samples and surpasses OpenAI’s o1 and GPT-4o models on the out-of-distribution (OOD) V* Bench. |
Direct Multi-Turn Preference Optimization for Language Agents (2024.emnlp-main)
Copied to clipboard
| Challenge: | Extensive experiments on three multi-turn agent task datasets confirm the effectiveness and superiority of the DMPO loss function. |
| Approach: | They propose a novel loss function for multi-turn agent tasks that replaces the policy constraint with the state-action occupancy measure constraint and adds length normalization to the Bradley-Terry model. |
| Outcome: | Experiments on three multi-turn agent task datasets confirm the effectiveness and superiority of the proposed loss function. |