Large language models have made strong reasoning gains through supervised fine-tuning, reinforcement learning, and on-policy distillation, yet these post-training methods are usually evaluated only by final-answer accuracy. We study how they reshape confidence during reasoning. We introduce a three-stage calibration framework that evaluates confidence before, during, and after chain-of-thought generation, corresponding to difficulty estimation, early termination, and answer aggregation. Through a controlled comparison on mathematical reasoning benchmarks, we find that OPD provides the most useful pre-reasoning confidence, SFT gives the strongest online signal for early stopping, and RL produces the most reliable trace-level signal for aggregation. We further show that confidence reliability is position-dependent: RL confidence becomes informative after a path-commitment phase, while OPD confidence is useful early but can become inversely calibrated later. Based on this observation, we propose PosConf, a position-aware confidence strategy that uses confidence only from reliable relative-position intervals. PosConf improves RL answer aggregation by 6.1 points over majority voting and consistently improves OPD early stopping under tight token budgets, with gains up to 4.3 points by avoiding its later inverse-calibration region, showing that confidence in reasoning models should be used both stage-wise and position-awarely. Our code is available at https://github.com/EIT-NLP/Post-Training-Calibration.

Large Language Models (LLMs) trained on massive data capture rich information embedded in the training data. However, this also introduces the risk of privacy leakage, particularly involving personally identifiable information (PII). Although previous studies have shown that this risk can be mitigated through methods such as privacy neurons, they all assume that both the (sensitive) training data and user queries are in English. We show that they cannot defend against the privacy leakage in cross-lingual contexts: even if the training data is exclusively in one language, these (private) models may still reveal private information when queried in another language. In this work, we first investigate the information flow of cross-lingual privacy leakage to give a better understanding. We find that LLMs process private information in the middle layers, where representations are largely shared across languages. The risk of leakage peaks when converted to a language-specific space in later layers. Based on this, we identify privacy-universal neurons and language-specific privacy neurons. Privacy-universal neurons influence privacy leakage across all languages, while language-specific privacy neurons are only related to specific languages. By deactivating these neurons, the cross-lingual privacy leakage risk is reduced by 23.3%-31.6%.

Regret in large language models (LLMs) refers to their explicit expression of regret when confronted with evidence contradicting previously generated misinformation. Understanding the neural encoding of regret is crucial for advancing artificial metacognition and model reliability. However, three key gaps remain: the absence of specialized datasets for regret expressions, the lack of understanding of how early regret signals emerge in hidden states, and the unknown neuron-level architecture underlying regret encoding. To address these gaps, we make three contributions: (1) we propose a multi-stage workflow for constructing the first regret dataset and train probes achieving near-perfect accuracy across LLaMA-2-7B, Mistral-7B, and Vicuna-13B. (2) we discover that regret signals begin emerging in preceding hidden states before the regret token appears, and that probes generalize to semantically related keywords such as "incorrect" and "apologize." (3) at the neuron level, we introduce the Regret Dominance Score (RDS) and Group Impact Coefficient (GIC) to reveal a compositional regret architecture where cooperative neuron groups collectively encode regret, with larger models (70B) exhibiting more pronounced compositional effects. Targeted interventions on these neurons successfully suppress regret expression in generated text, validating our mechanistic understanding. Our code and data are available at https://github.com/AbnerAI/LLMs-Regret-Before-They-Say-It.