Papers with LLM agents
LongSpec: Long-Context Lossless Speculative Decoding with Efficient Drafting and Verification (2026.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) can process extremely long contexts, requiring efficient inference over extended inputs. |
| Approach: | They propose a model that uses a constant-sized key-value cache to train long-context models. |
| Outcome: | Experimental results show that LongSpec achieves 3.26x speedup over strong Flash Attention baselines and 2.34x wall clock time on four math reasoning tasks. |
Beyond Blind Following: Evaluating Robustness of LLM Agents under Imperfect Guidance (2026.eacl-long)
Copied to clipboard
Yao Fu, Ran Qiu, Xinhe Wang, Jacob Sansom, Sathvika Ayyappa Prabhu, Huijie Tang, Jaekyeom Kim, Sungryull Sohn, Honglak Lee
| Challenge: | Large language models (LLMs) have shown strong capabilities as task-solving agents across interactive domains, but in complex environments, auxiliary guidance may be imperfect. |
| Approach: | They propose a benchmark to measure the robustness of large language models under imperfect guidance. |
| Outcome: | The proposed benchmark compared LLM agents in navigation, cooking, and gaming in a variety of environments with auxiliary guidance and noisy or underspecified instructions extracted from demonstrations. |
No More Stale Feedback: Co-Evolving Critics for Open-World Agent Learning (2026.acl-long)
Copied to clipboard
Zhicong Li, Lingjie Jiang, Yulan Hu, Xingchen Zeng, Yixia Li, Xiangwen Zhang, Guanhua Chen, Zheng Pan, Xin Li, Yong Liu
| Challenge: | Current methods for training Large Language Model agents rely on static or offline critic models, which fail to adapt as the policy evolves. |
| Approach: | They propose a framework that integrates a critique and a policy to optimize the policy and critic through a synchronized co-evolutionary loop. |
| Outcome: | The proposed framework yields more stable training and higher long-horizon task success across open-world environments. |
RealSafe: Quantifying Safety Risks of Language Agents in Real-World (2025.coling-main)
Copied to clipboard
| Challenge: | Recent advances in Large Language Models (LLMs) have attracted much attention in Artificial Intelligence. |
| Approach: | They propose a framework to rigorously assess the safety and reliability of large language model (LLM) agents in real application scenarios. |
| Outcome: | The framework evaluates the safety and reliability of large language model (LLM) agents in 14 different application scenarios utilizing three contexts - standard operations, ambiguous interactions, and malicious behaviors. |
Planning with Multi-Constraints via Collaborative Language Agents (2025.coling-main)
Copied to clipboard
| Challenge: | Recent advances in neural language models have sparked a new surge of intelligent agent research. |
| Approach: | They propose a method for collaborative LLM-based multi-agent systems that simplifies complex task planning with constraints by decomposing it into a hierarchy of subordinate tasks. |
| Outcome: | The proposed method achieves an average success rate of 42.68% on two constraint-intensive benchmarks, TravelPlanner and API-Bank. |
Contextual Experience Replay for Self-Improvement of Language Agents (2025.acl-long)
Copied to clipboard
| Challenge: | Existing large language model (LLM) agents fail in complex tasks without any environment-specific experiences. |
| Approach: | They propose a framework that accumulates and synthesizes past experiences into a dynamic memory buffer to enable efficient self-improvement for language agents in their context window. |
| Outcome: | The proposed framework improves performance on WebArena and VisualWebAren . it surpasses tree search method with fewer token costs and achieves state-of-the-art performance of 31.9%. |
PGPO: Enhancing Agent Reasoning via Pseudocode-style Planning Guided Preference Optimization (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing LLM agents generate verbose and inefficient natural language plans to guide reasoning, which restricts agents’ ability to generalize across similar tasks. |
| Approach: | They propose a pseudocode-style planning guide optimization method that captures the structural logic of reasoning and uses two planning-oriented rewards to enhance agent learning. |
| Outcome: | The proposed method outperforms existing LLM agents on representative agent benchmarks and outperformed the current leading baselines. |
CAR: Empowering Agents with Dynamic Tool Synthesis and Global Trajectory Rectification (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing LLM agents are brittle in open-ended environments due to two limitations: 1) a closed action space; 2) myopic error recovery. |
| Approach: | They propose a novel architecture that augments the action space and revises global strategies by adding a reflective replanning mechanism to the system. |
| Outcome: | Experiments show that CAR outperforms baselines in a diagnostic benchmark with pruned toolsets to simulate tool scarcity. |
Consolidation or Adaptation? PRISM: Disentangling SFT and RL Data via Gradient Concentration (2026.acl-long)
Copied to clipboard
Yang Zhao, Yangou Ouyang, Xiao Ding, Hepeng Wang, Bibo Cai, Kai Xiong, Jinglong Gao, Zhouhao Sun, Li Du, Bing Qin, Ting Liu
| Challenge: | Existing data arbitration strategies for large language model training rely on surface-level heuristics that fail to diagnose intrinsic learning needs. |
| Approach: | They propose a framework that arbitrates data based on its degree of cognitive conflict with the model's existing knowledge. |
| Outcome: | Extensive experiments on WebShop and ALFWorld show that PRISM outperforms state-of-the-art hybrid methods while reducing computational costs by up to 3.22 . |
Learning to Ask: When LLM Agents Meet Unclear Instruction (2025.emnlp-main)
Copied to clipboard
Wenxuan Wang, Shi Juluan, Zixuan Ling, Yuk-Kit Chan, Chaozheng Wang, Cheryl Lee, Youliang Yuan, Jen-tse Huang, Wenxiang Jiao, Michael R. Lyu
| Challenge: | Despite their impressive capabilities, LLMs struggle with complex computations and delivering accurate, timely information. |
| Approach: | They propose a framework that prompts LLM agents to ask questions when they encounter obstacles due to unclear instructions and an automated evaluation tool called ToolEvaluator. |
| Outcome: | The proposed framework outperforms existing frameworks for tool learning in the Noisy ToolBench. |
PersonaAgent: Bridging Memory and Action for Personalized LLM Agents (2026.findings-acl)
Copied to clipboard
Weizhi Zhang, Xinyang Zhang, Chenwei Zhang, Liangwei Yang, Jingbo Shang, Zhepei Wei, Henry Peng Zou, Zijie Huang, Zhengyang Wang, Yifan Gao, Xiaoman Pan, Lian Xiong, Jingguo Liu, Philip S. Yu, Xian Li
| Challenge: | Existing Large Language Model (LLM) enabled agents lack flexibility to respond to users’ varying needs and preferences. |
| Approach: | They propose a test-time user-preference alignment strategy that optimizes the persona prompt, ensuring real-time preference alignment through textual loss feedback between simulated and ground-truth responses. |
| Outcome: | The proposed framework outperforms baseline methods in real-time and in real applications. |
PEARL: Self-Evolving Assistant for Time Management with Reinforcement Learning (2026.acl-long)
Copied to clipboard
| Challenge: | Scheduling logistics associated with meetings can amount to hours each week . a lack of automation can cause a lot of errors and can be costly . |
| Approach: | They propose a benchmark for long-horizon calendar conflict resolution that automates calendar conflict management . they propose 'pearl' that augments the language agent with an external preference memory . |
| Outcome: | a new benchmark for long-horizon calendar conflict resolution shows that existing agents perform poorly with high error rates. |
InfiAgent: An Infinite-Horizon Framework for General-Purpose Autonomous Agents (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing LLMs break down on long-horizon tasks due to unbounded context growth and accumulated errors. |
| Approach: | They propose a framework that externalizes persistent state into a file-centric state abstraction and keeps the agent’s reasoning context strictly bounded regardless of task duration. |
| Outcome: | Experiments on DeepResearch and an 80-paper literature review show that the proposed framework maintains higher long-horizon coverage than baseline models without task-specific fine-tuning. |