LSTM-MAS: A Long Short-Term Memory Inspired Multi-Agent System for Long-Context Understanding
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908981600976896 |
|---|---|
| author | Jiang, Yichen Yuan, Jiakang Tu, Chongjun Ye, Peng Chen, Tao |
| author_facet | Jiang, Yichen Yuan, Jiakang Tu, Chongjun Ye, Peng Chen, Tao |
| contents | Effectively processing long contexts remains a fundamental yet unsolved challenge for large language models (LLMs). Existing single-LLM-based methods primarily reduce the context window or optimize the attention mechanism, but they often encounter additional computational costs or constrained expanded context length. While multi-agent-based frameworks can mitigate these limitations, they remain susceptible to the accumulation of errors and the propagation of hallucinations. In this work, we draw inspiration from the Long Short-Term Memory (LSTM) architecture to design a Multi-Agent System called LSTM-MAS, emulating LSTM's hierarchical information flow and gated memory mechanisms for long-context understanding. Specifically, LSTM-MAS organizes agents in a chained architecture, where each node comprises a worker agent for segment-level comprehension, a filter agent for redundancy reduction, a judge agent for continuous error detection, and a manager agent for globally regulates information propagation and retention, analogous to LSTM and its input gate, forget gate, constant error carousel unit, and output gate. These novel designs enable controlled information transfer and selective long-term dependency modeling across textual segments, which can effectively avoid error accumulation and hallucination propagation. We conducted an extensive evaluation of our method. Compared with the previous best multi-agent approach, CoA, our model achieves improvements of 97.97%, 65.75%, 122.19%, 39.61% and 10.80% on Narrative QA, Qasper, HotpotQA, 2WikiMQA and MuSiQue, respectively. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_11913 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LSTM-MAS: A Long Short-Term Memory Inspired Multi-Agent System for Long-Context Understanding Jiang, Yichen Yuan, Jiakang Tu, Chongjun Ye, Peng Chen, Tao Computation and Language Artificial Intelligence 68T42 I.2.11 Effectively processing long contexts remains a fundamental yet unsolved challenge for large language models (LLMs). Existing single-LLM-based methods primarily reduce the context window or optimize the attention mechanism, but they often encounter additional computational costs or constrained expanded context length. While multi-agent-based frameworks can mitigate these limitations, they remain susceptible to the accumulation of errors and the propagation of hallucinations. In this work, we draw inspiration from the Long Short-Term Memory (LSTM) architecture to design a Multi-Agent System called LSTM-MAS, emulating LSTM's hierarchical information flow and gated memory mechanisms for long-context understanding. Specifically, LSTM-MAS organizes agents in a chained architecture, where each node comprises a worker agent for segment-level comprehension, a filter agent for redundancy reduction, a judge agent for continuous error detection, and a manager agent for globally regulates information propagation and retention, analogous to LSTM and its input gate, forget gate, constant error carousel unit, and output gate. These novel designs enable controlled information transfer and selective long-term dependency modeling across textual segments, which can effectively avoid error accumulation and hallucination propagation. We conducted an extensive evaluation of our method. Compared with the previous best multi-agent approach, CoA, our model achieves improvements of 97.97%, 65.75%, 122.19%, 39.61% and 10.80% on Narrative QA, Qasper, HotpotQA, 2WikiMQA and MuSiQue, respectively. |
| title | LSTM-MAS: A Long Short-Term Memory Inspired Multi-Agent System for Long-Context Understanding |
| topic | Computation and Language Artificial Intelligence 68T42 I.2.11 |
| url | https://arxiv.org/abs/2601.11913 |