Function calling enables large language models (LLMs) to interact with external systems by leveraging tools and APIs. When faced with multi-step tool usage, LLMs still struggle with tool selection, parameter generation, and tool-chain planning. Existing methods typically rely on manually designing task-specific demonstrations, or retrieving from a curated library. These approaches demand substantial expert effort and prompt engineering becomes increasingly complex and inefficient as tool diversity and task difficulty scale. To address these challenges, we propose a self-guided method, Stepwise Experience Recall (SEER), which performs fine-grained, stepwise retrieval from a continually updated experience pool. Instead of relying on static or manually curated library, SEER incrementally augments the experience pool with past successful trajectories, enabling continuous expansion of the pool and improved model performance over time. Evaluated on the ToolQA benchmark, SEER achieves an average improvement of 6.1% on easy and 4.7% on hard questions. We further test SEER on $τ$-bench, which includes two real-world domains. Powered by Qwen2.5-7B and Qwen2.5-72B models, SEER demonstrates substantial accuracy gains of 7.44% and 23.38%, respectively.

Recent advances in large language models (LLMs) have highlighted the potential of reinforcement learning with verifiable rewards (RLVR) to enhance reasoning capabilities through extended output sequences. However, traditional RL frameworks face inefficiencies when handling ultra-long outputs due to long-tail sequence distributions and entropy collapse during training. To address these challenges, we propose an Ultra-Long Output Reinforcement Learning (UloRL) approach for advancing large language models' reasoning abilities. Specifically, we divide ultra long output decoding into short segments, enabling efficient training by mitigating delays caused by long-tail samples. Additionally, we introduce dynamic masking of well-Mastered Positive Tokens (MPTs) to prevent entropy collapse. Experimental results demonstrate the effectiveness of our approach. On the Qwen3-30B-A3B model, RL with segment rollout achieved 2.06x increase in training speed, while RL training with 128k-token outputs improves the model's performance on AIME2025 from 70.9\% to 85.1\% and on BeyondAIME from 50.7\% to 61.9\%, even surpassing Qwen3-235B-A22B with remarkable gains. These findings underscore the potential of our methods to advance the reasoning capabilities of LLMs with ultra-long sequence generation. We will release our code and model for further use by the community.

Accessing knowledge via multilingual natural-language interfaces is one of the emerging challenges in the field of information retrieval and related ones. Structured knowledge stored in knowledge graphs can be queried via a specific query language (e.g., SPARQL). Therefore, one needs to transform natural-language input into a query to fulfill an information need. Prior approaches mostly focused on combining components (e.g., rule-based or neural-based) that solve downstream tasks and come up with an answer at the end. We introduce mKGQAgent, a human-inspired framework that breaks down the task of converting natural language questions into SPARQL queries into modular, interpretable subtasks. By leveraging a coordinated LLM agent workflow for planning, entity linking, and query refinement - guided by an experience pool for in-context learning - mKGQAgent efficiently handles multilingual KGQA. Evaluated on the DBpedia- and Corporate-based KGQA benchmarks within the Text2SPARQL challenge 2025, our approach took first place among the other participants. This work opens new avenues for developing human-like reasoning systems in multilingual semantic parsing.

Reinforcement learning (RL) for large language models is an energy-intensive endeavor: training can be unstable, and the policy may gradually drift away from its pretrained weights. We present \emph{RLEP}\, -- \,Reinforcement Learning with Experience rePlay\, -- \,a two-phase framework that first collects verified trajectories and then replays them during subsequent training. At every update step, the policy is optimized on mini-batches that blend newly generated rollouts with these replayed successes. By replaying high-quality examples, RLEP steers the model away from fruitless exploration, focuses learning on promising reasoning paths, and delivers both faster convergence and stronger final performance. On the Qwen2.5-Math-7B base model, RLEP reaches baseline peak accuracy with substantially fewer updates and ultimately surpasses it, improving accuracy on AIME-2024 from 38.2% to 39.9%, on AIME-2025 from 19.8% to 22.3%, and on AMC-2023 from 77.0% to 82.2%. Our code, datasets, and checkpoints are publicly available at https://github.com/Kwai-Klear/RLEP to facilitate reproducibility and further research.

With the rapid development of LLM-based agents, increasing attention has been given to their social interaction and strategic reasoning capabilities. However, existing Werewolf-based benchmarking platforms suffer from overly simplified game settings, incomplete evaluation metrics, and poor scalability. To address these limitations, we propose WereWolf-Plus, a multi-model, multi-dimensional, and multi-method benchmarking platform for evaluating multi-agent strategic reasoning in the Werewolf game. The platform offers strong extensibility, supporting customizable configurations for roles such as Seer, Witch, Hunter, Guard, and Sheriff, along with flexible model assignment and reasoning enhancement strategies for different roles. In addition, we introduce a comprehensive set of quantitative evaluation metrics for all special roles, werewolves, and the sheriff, and enrich the assessment dimensions for agent reasoning ability, cooperation capacity, and social influence. WereWolf-Plus provides a more flexible and reliable environment for advancing research on inference and strategic interaction within multi-agent communities. Our code is open sourced at https://github.com/MinstrelsyXia/WereWolfPlus.

Large Language Model-based multi-agent systems (MAS) have shown remarkable progress in solving complex tasks through collaborative reasoning and inter-agent critique. However, existing approaches typically treat each task in isolation, resulting in redundant computations and limited generalization across structurally similar tasks. To address this, we introduce multi-agent cross-task experiential learning (MAEL), a novel framework that endows LLM-driven agents with explicit cross-task learning and experience accumulation. We model the task-solving workflow on a graph-structured multi-agent collaboration network, where agents propagate information and coordinate via explicit connectivity. During the experiential learning phase, we quantify the quality for each step in the task-solving workflow and store the resulting rewards along with the corresponding inputs and outputs into each agent's individual experience pool. During inference, agents retrieve high-reward, task-relevant experiences as few-shot examples to enhance the effectiveness of each reasoning step, thereby enabling more accurate and efficient multi-agent collaboration. Experimental results on diverse datasets demonstrate that MAEL empowers agents to learn from prior task experiences effectively-achieving faster convergence and producing higher-quality solutions on current tasks.

Large language models (LLMs) can perform complex reasoning by generating intermediate thoughts under zero-shot or few-shot settings. However, zero-shot prompting always encounters low performance, and the superior performance of few-shot prompting hinges on the manual-crafted demonstrations. In this paper, we present RoSE (Reasoning with Orchestrated Streaming Experiences), a general framework for solving reasoning tasks that can self-improve without complex external efforts. To enable RoSE, we describe an architecture that extends an LLM to store all answered questions and their thoughts in a streaming experience pool then orchestrates helpful questions from the pool to assist in answering new questions. To set up a question-aware orchestration mechanism, RoSE first calculates the similarity of each question in the pool with a new test question. Since the solution to each answered question is not always correct, RoSE will sort the questions according to their similarity with the new question, and then uniformly divide them into multiple buckets. It finally extracts one question from each bucket to make these extracted questions more diverse. To make these extracted questions help RoSE answer new questions as much as possible, we introduce two other attributes of uncertainty and complexity for each question. RoSE will preferentially select the questions with low uncertainty and high complexity from each bucket. We evaluate the versatility of RoSE in various reasoning tasks, LLMs, and CoT methods.

Large language model agents have demonstrated remarkable advancements across various complex tasks. Recent works focus on optimizing the agent team or employing self-reflection to iteratively solve complex tasks. Since these agents are all based on the same LLM, only conducting self-evaluation or removing underperforming agents does not substantively enhance the capability of the agents. We argue that a comprehensive evaluation and accumulating experience from evaluation feedback is an effective approach to improving system performance. In this paper, we propose Reusable Experience Accumulation with 360$^\circ$ Assessment (360$^\circ$REA), a hierarchical multi-agent framework inspired by corporate organizational practices. The framework employs a novel 360$^\circ$ performance assessment method for multi-perspective performance evaluation with fine-grained assessment. To enhance the capability of agents in addressing complex tasks, we introduce dual-level experience pool for agents to accumulate experience through fine-grained assessment. Extensive experiments on complex task datasets demonstrate the effectiveness of 360$^\circ$REA.

Autonomous agents powered by large language models (LLMs) show significant potential for achieving high autonomy in various scenarios such as software development. Recent research has shown that LLM agents can leverage past experiences to reduce errors and enhance efficiency. However, the static experience paradigm, reliant on a fixed collection of past experiences acquired heuristically, lacks iterative refinement and thus hampers agents' adaptability. In this paper, we introduce the Iterative Experience Refinement framework, enabling LLM agents to refine experiences iteratively during task execution. We propose two fundamental patterns: the successive pattern, refining based on nearest experiences within a task batch, and the cumulative pattern, acquiring experiences across all previous task batches. Augmented with our heuristic experience elimination, the method prioritizes high-quality and frequently-used experiences, effectively managing the experience space and enhancing efficiency. Extensive experiments show that while the successive pattern may yield superior results, the cumulative pattern provides more stable performance. Moreover, experience elimination facilitates achieving better performance using just 11.54% of a high-quality subset.

Prioritized Experience Replay (PER) is a technical means of deep reinforcement learning by selecting experience samples with more knowledge quantity to improve the training rate of neural network. However, the non-uniform sampling used in PER inevitably shifts the state-action space distribution and brings the estimation error of Q-value function. In this paper, an Attention Loss Adjusted Prioritized (ALAP) Experience Replay algorithm is proposed, which integrates the improved Self-Attention network with Double-Sampling mechanism to fit the hyperparameter that can regulate the importance sampling weights to eliminate the estimation error caused by PER. In order to verify the effectiveness and generality of the algorithm, the ALAP is tested with value-function based, policy-gradient based and multi-agent reinforcement learning algorithms in OPENAI gym, and comparison studies verify the advantage and efficiency of the proposed training framework.