Answer Probing-Guided Search for Diverse Solution Exploration of LLMs
2026-08-31 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors address the problem that Large Language Models (LLMs) often give only one strong answer instead of many different possible solutions. They discover that typical methods using language similarity struggle to tell apart truly different solutions. To fix this, the authors propose "Answer Probing," which looks deeper into the model's reasoning by checking the hidden states and confidence of potential answers. They use this to guide a better search method called Answer Probing-Guided Tree Search (APTS), which leads to more diverse and reliable solutions across several reasoning tests.
Large Language ModelsSolution DiversityTree SearchSemantic EmbeddingsHidden StatesPerplexityReasoning PathsAnswer ProbingInferenceModel Robustness
Authors
Yi Fang, Que Shen, Chengpeng Li, Boyi Deng, Wei Shi, Wenjie Wang, Fuli Feng, Fengli Xu, Dayiheng Liu
Abstract
Generating multiple diverse and high-quality solutions is valuable for many applications, such as code-test generation and drug discovery. However, Large Language Models (LLMs) tend to converge on a single high-confidence solution during inference, limiting exploration of alternative valid solution paths. Existing test-time methods promote diversity through tree-like search and prune semantically similar branches using response-level semantic embeddings. However, we find that such embeddings are easily confounded by linguistic and stylistic similarities, making it difficult to distinguish genuinely distinct solution paths. To address this, we introduce Answer Probing, which probes the potential answer an LLM would reach from an intermediate reasoning path. We demonstrate that the hidden states of probed answers more effectively differentiate distinct solution paths than semantic embeddings, and the perplexity of probed answers serves as a practical proxy for reasoning correctness. Based on these findings, we propose Answer Probing-Guided Tree Search (APTS), which guides the tree search by the probed answers' hidden state similarity and perplexity. Experiments on three reasoning tasks across two LLMs show that APTS consistently enhances solution diversity, demonstrating its effectiveness and robustness.