LLMs Can Generate a Better Answer by Aggregating Their Own Responses
This addresses the limitation of LLMs in handling complex problems without requiring discriminative capabilities, offering a more general solution for open-ended tasks, though it is incremental as it builds on existing aggregation ideas.
The paper tackles the problem of LLMs performing poorly on complex tasks when using self-correction or response selection methods by proposing Generative Self-Aggregation (GSA), a prompting method that aggregates multiple diverse responses to synthesize an improved answer, resulting in effective quality improvements across tasks like mathematical reasoning and code synthesis.
Large Language Models (LLMs) have shown remarkable capabilities across tasks, yet they often require additional prompting techniques when facing complex problems. While approaches like self-correction and response selection have emerged as popular solutions, recent studies have shown these methods perform poorly when relying on the LLM itself to provide feedback or selection criteria. We argue this limitation stems from the fact that common LLM post-training procedures lack explicit supervision for discriminative judgment tasks. In this paper, we propose Generative Self-Aggregation (GSA), a novel prompting method that improves answer quality without requiring the model's discriminative capabilities. GSA first samples multiple diverse responses from the LLM, then aggregates them to obtain an improved solution. Unlike previous approaches, our method does not require the LLM to correct errors or compare response quality; instead, it leverages the model's generative abilities to synthesize a new response based on the context of multiple samples. While GSA shares similarities with the self-consistency (SC) approach for response aggregation, SC requires specific verifiable tokens to enable majority voting. In contrast, our approach is more general and can be applied to open-ended tasks. Empirical evaluation demonstrates that GSA effectively improves response quality across various tasks, including mathematical reasoning, knowledge-based problems, and open-ended generation tasks such as code synthesis and conversational responses.