Can Memory-Augmented Language Models Generalize on Reasoning-in-a-Haystack Tasks?
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866917950956503040 |
|---|---|
| author | Das, Payel Ko, Ching-Yun Dai, Sihui Kollias, Georgios Chaudhury, Subhajit Lozano, Aurelie |
| author_facet | Das, Payel Ko, Ching-Yun Dai, Sihui Kollias, Georgios Chaudhury, Subhajit Lozano, Aurelie |
| contents | Large language models often expose their brittleness in reasoning tasks, especially while executing long chains of reasoning over context. We propose MemReasoner, a new and simple memory-augmented LLM architecture, in which the memory learns the relative order of facts in context, and enables hopping over them, while the decoder selectively attends to the memory. MemReasoner is trained end-to-end, with optional supporting fact supervision of varying degrees. We train MemReasoner, along with existing memory-augmented transformer models and a state-space model, on two distinct synthetic multi-hop reasoning tasks. Experiments performed under a variety of challenging scenarios, including the presence of long distractor text or target answer changes in test set, show strong generalization of MemReasoner on both single- and two-hop tasks. This generalization of MemReasoner is achieved using none-to-weak supporting fact supervision (using none and 1\% of supporting facts for one- and two-hop tasks, respectively). In contrast, baseline models overall struggle to generalize and benefit far less from using full supporting fact supervision. The results highlight the importance of explicit memory mechanisms, combined with additional weak supervision, for improving large language model's context processing ability toward reasoning tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_07903 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Can Memory-Augmented Language Models Generalize on Reasoning-in-a-Haystack Tasks? Das, Payel Ko, Ching-Yun Dai, Sihui Kollias, Georgios Chaudhury, Subhajit Lozano, Aurelie Computation and Language Machine Learning Large language models often expose their brittleness in reasoning tasks, especially while executing long chains of reasoning over context. We propose MemReasoner, a new and simple memory-augmented LLM architecture, in which the memory learns the relative order of facts in context, and enables hopping over them, while the decoder selectively attends to the memory. MemReasoner is trained end-to-end, with optional supporting fact supervision of varying degrees. We train MemReasoner, along with existing memory-augmented transformer models and a state-space model, on two distinct synthetic multi-hop reasoning tasks. Experiments performed under a variety of challenging scenarios, including the presence of long distractor text or target answer changes in test set, show strong generalization of MemReasoner on both single- and two-hop tasks. This generalization of MemReasoner is achieved using none-to-weak supporting fact supervision (using none and 1\% of supporting facts for one- and two-hop tasks, respectively). In contrast, baseline models overall struggle to generalize and benefit far less from using full supporting fact supervision. The results highlight the importance of explicit memory mechanisms, combined with additional weak supervision, for improving large language model's context processing ability toward reasoning tasks. |
| title | Can Memory-Augmented Language Models Generalize on Reasoning-in-a-Haystack Tasks? |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2503.07903 |