Lift Yourself Up: Retrieval-augmented Text Generation with Self Memory

TL;DR

Selfmem introduces a retrieval-augmented generation framework that uses its own generated outputs as unbounded memory to iteratively enhance generation quality. By coupling a retrieval-augmented generator with a memory selector and leveraging two modes of generation, Selfmem creates a feedback loop (primal-dual) where better memory improves generation and better generation yields better memory. Across machine translation, abstractive summarization, and dialogue, Selfmem achieves state-of-the-art results on several datasets and provides detailed analyses to identify bottlenecks and avenues for future work. This framework offers a practical pathway to surpass fixed-memory limits in retrieval-augmented models and informs memory-driven improvements for both small models and few-shot LLM setups.

Abstract

With direct access to human-written reference as memory, retrieval-augmented generation has achieved much progress in a wide range of text generation tasks. Since better memory would typically prompt better generation~(we define this as primal problem). The traditional approach for memory retrieval involves selecting memory that exhibits the highest similarity to the input. However, this method is constrained by the quality of the fixed corpus from which memory is retrieved. In this paper, by exploring the duality of the primal problem: better generation also prompts better memory, we propose a novel framework, selfmem, which addresses this limitation by iteratively employing a retrieval-augmented generator to create an unbounded memory pool and using a memory selector to choose one output as memory for the subsequent generation round. This enables the model to leverage its own output, referred to as self-memory, for improved generation. We evaluate the effectiveness of selfmem on three distinct text generation tasks: neural machine translation, abstractive text summarization, and dialogue generation, under two generation paradigms: fine-tuned small model and few-shot LLM. Our approach achieves state-of-the-art results in four directions in JRC-Acquis, XSum (50.3 ROUGE-1), and BigPatent (62.9 ROUGE-1), demonstrating the potential of self-memory in enhancing retrieval-augmented generation models. Furthermore, we conduct thorough analyses of each component in the selfmem framework to identify bottlenecks and provide insights for future research.

Figures (7)

• Figure 1: Relation between memory and hypothesis on JRC-Acquis En$\rightarrow$De dataset. The hypothesis is generated by a retrieval-augmented translator whose memory is retrieved from the training set. The X-axis represents the similarity between memory and the reference.
• Figure 2: Overall framework. There are two components in Selfmem, a retrieval-augmented generator (a) and a memory selector (b). For the primal problem, (a) takes source and memory as input to generate candidates for (b). For the dual problem, (b) takes as input source and generated candidates to select memory for (a).
• Figure 3: (a) shows generation quality in the iteration process with different $S_{\theta}$ in both trainable generator architectures. (b) shows candidates quality in the iteration process with an oracle $S_{\theta}$.
• Figure 4: 1-gram F1 score sorted by training corpus frequency.
• Figure 5: Results of summarization task on XSum and BigPatent measured by ROUGE.