Tree of Thoughts (ToT) reasoning is fundamentally bottlenecked by its generate-then-score paradigm, which decodes every candidate branch into text and evaluates it through a decoupled scoring stage. Prior efficiency efforts primarily shrink the search breadth, leaving this prohibitive per-step textual evaluation intact. To resolve this, we propose ITOT (Implicit Tree of Thoughts), a training-free framework that recasts tree-structured reasoning as a continuous search within a model’s latent space. In this framework, branch quality is assessed intrinsically from hidden states via alignment with a contrastive prototype distilled from successful and failed traces. This elegantly collapses explicit branch enumeration and textual scoring into a single latent generate-evaluate-decide loop. Across six reasoning benchmarks and five LLM backbones, ITOT consistently improves mean accuracy over both the strongest tree-search and single-path baselines by roughly ten absolute points. Crucially, it achieves this while cutting token consumption by nearly an order of magnitude relative to explicit tree-search methods, operating even below the token budget of vanilla Chain of Thought. We will release our code to facilitate further research.

The most recent research uses reinforcement learning (RL) to post-train Multi-modal Large Language Models (MLLMs) such that these models are able to iteratively call search engines to dynamically access external knowledge when handling complex Visual Question Answering (VQA) tasks. However, existing methods face two major limitations in effectiveness and efficiency: i) For effectiveness, the objective of these methods, which only considers the correctness of the generated final response, overlooks the quality of intermediate search results, thus leading to suboptimal search strategies. ii) For efficiency, existing methods often unnecessarily invoke search calls during reasoning, making the inference inefficient. To address these issues, we propose , a customized dual-objective reinforcement learning framework to improve the search strategies of MLLMs, enhancing their search quality yet minimizing search frequency. The key ideas include (1) a reward function that promotes correct reasoning trajectories with fewer search calls; and (2) dual optimization objectives that jointly optimize search quality and answer correctness. Extensive experiments on 3 real-world datasets demonstrate that DORA outperforms state-of-the-art methods, achieving up to 8.4% higher accuracy while reducing the number of search calls by 9.7%.

Hallucinations arise when large language models (LLMs) guess rather than acknowledge their underlying uncertainty. Existing static strategies for mitigating hallucinations have been only partially successful, largely because they do not explicitly model the information gain from interacting with the external environment. Researchers need a general method to proactively steer users toward informative clarifications, thereby unlocking the model’s effective capacity under underspecified inputs. We model the uncertainty of LLMs in interactive settings and uncover the mechanism of active calibration between model concepts and human evaluations, improving reliability. We show that calibration error in LLMs density estimation admits a non-vanishing lower bound under non-interactive learning, while interaction empirically reduces it. We further characterize that calibration error identifies informative queries and that calibration can be accelerated by shifting query distributions from imbalanced to balanced regimes. Guided by these insights, we propose a calibration-driven Interactive Learning Strategy (ILS) that selects clarification queries by optimizing calibration error, providing both theoretical guarantees and empirical gains for reliability. Code and data are available at https://github.com/zhouyeah215/Demystifying_Uncertainty.