Lune

ACL Findings2026

Agent-Dice: Disentangling Knowledge Updates via Geometric Consensus for Agent Continual Learning

Zheng Wu, Xingyu Lou, Xinbei Ma, Yansi Li, Weiwen Liu, Weinan Zhang, Jun Wang, Zhuosheng Zhang

Abstract

Large Language Model (LLM)-based agents significantly extend the utility of LLMs by interacting with dynamic environments. However, enabling agents to continually learn new tasks without catastrophic forgetting remains a critical challenge, known as the stability-plasticity dilemma. In this work, we argue that this dilemma fundamentally arises from the failure to explicitly distinguish between common knowledge shared across tasks and conflicting knowledge introduced by task-specific interference. To address this, we propose Agent-Dice, a parameter fusion framework based on directional consensus evaluation. Concretely, Agent-Dice disentangles knowledge updates through a twostage process: geometric consensus filtering to prune conflicting gradients, and curvaturebased importance weighting to amplify shared semantics. We provide a rigorous theoretical analysis that establishes the validity of the proposed fusion scheme and offers insight into the origins of the stability-plasticity dilemma. Extensive experiments on GUI agents and tooluse agent domains demonstrate that Agent-Dice exhibits outstanding continual learning performance with minimal computational overhead and parameter updates.

The codes are available at https://github.com/ Wuzheng02/Agent-Dice.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

1 top-tier paper has cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs cites Toolformer: Language Models Can Teach Themselves to Use Tools
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers1

  1. LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward DecompositionYanyu Chen, Jiyue Jiang, Dianzhi Yu, Zheng Wu et al.KDD 2026

Builds on24

  1. Toolformer: Language Models Can Teach Themselves to Use ToolsTimo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu et al.NeurIPS 2023 · 5,989 citations
  2. ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIsYujia Qin, Shihao Liang, Yining Ye, Kunlun Zhu et al.ICLR 2024 · 1,469 citations
  3. Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference timeMitchell Wortsman, Gabriel Ilharco, Samir Yitzhak Gadre, Rebecca Roelofs et al.ICML 2022 · 1,464 citations
  4. A-Mem: Agentic Memory for LLM AgentsWujiang Xu, Zujie Liang, Kai Mei, Hang Gao et al.NeurIPS 2025 · 1,138 citations
  5. Learning to Prompt for Continual LearningZifeng Wang, Zizhao Zhang, Chen-Yu Lee, Han Zhang et al.CVPR 2022 · 635 citations

Related papers

  1. DICE: Dynamic In-Context Example Selection in LLM Agents via Efficient Knowledge TransferRuoyu Wang, Junda Wu, Yu Xia, Tong Yu et al.KDD 2026 · 6 citations
  2. RECALL: REpresentation-aligned Catastrophic-forgetting ALLeviation via Hierarchical Model MergingBowen Wang, Haiyuan Wan, Liwen Shi, Chen Yang et al.EMNLP 2025
  3. GAM: Hierarchical Graph-based Agentic Memory for LLM AgentsZhaofen Wu, Hanrong Zhang, Fulin Lin, Wujiang Xu et al.ACL 2026 · 8 citations
  4. AdaDPI: Document-level Translation Adaptive Agent via Dynamic Parametric InternalizationHong Ren, Liting Deng, Shaolin Zhu, Deyi XiongACL 2026
  5. Recurrent Knowledge Identification and Fusion for Language Model Continual LearningYujie Feng, Xujia Wang, Zexin Lu, Shenghong Fu et al.ACL 2025
Dusk over the sea between two cliffs drawn in fine vertical lines