Skills help LLM agents reuse procedural knowledge in recurring workflows, but broad transferability alone does not guarantee workflow fit: a reusable skill may be semantically relevant yet fail under local tools, files, formats, and failure modes. We propose SkillMe, a workflow-situated skill evolution framework that adapts agent skills through repeated use in matched workflow recurrences. A Publisher induces and validates an initial skill from a successful execution, while Executors use the skill in agent-native sessions and provide trace-grounded feedback. SkillMe activates the matched skill on demand as workspace files, allowing agents to inspect and adopt guidance progressively without permanently installing it as a local skill. On the 71-task skill-needed subset of PinchBench with OpenClaw, SkillMe improves the overall mean score of all eight Executors over their paired no-skill baselines, with mean gains of +25.7 points. Comparisons show that evolved task-scoped skills are more stable than static skills and more reliable than broadly reusable skills. SkillMe thus provides an agent-native path to evolve reusable skills into effective workflow guidance through situated use.

Ensuring awareness of fairness and privacy in Large Language Models (LLMs) is critical. Interestingly, we discover a counter-intuitive trade-off phenomenon that enhancing an LLM’s privacy awareness through Supervised Fine-Tuning (SFT) methods significantly decreases its fairness awareness with thousands of samples. To address this issue, inspired by the information theory, we introduce a training-free method to Suppress the Privacy and faIrness coupled Neurons (SPIN), which theoretically and empirically decrease the mutual information between fairness and privacy awareness. Extensive experimental results demonstrate that SPIN eliminates the trade-off phenomenon and significantly improves LLMs’ fairness and privacy awareness simultaneously without compromising general capabilities, e.g., improving Qwen-2-7B-Instruct’s fairness awareness by 12.2% and privacy awareness by 14.0%.More crucially, SPIN remains robust and effective with limited annotated data or even when only malicious fine-tuning data is available, whereas SFT methods may fail to perform properly in such scenarios. Furthermore, we show that SPIN could generalize to other potential trade-off dimensions.We hope this study provides valuable insights into concurrently addressing fairness and privacy concerns in LLMs and can be integrated into comprehensive frameworks to develop more ethical and responsible AI systems. Our code is available at https://github.com/ChnQ/SPIN.

Ensuring the trustworthiness of large language models (LLMs) is crucial. Most studies concentrate on fully pre-trained LLMs to better understand and improve LLMs’ trustworthiness. In this paper, to reveal the untapped potential of pre-training, we pioneer the exploration of LLMs’ trustworthiness during this period, focusing on five key dimensions: reliability, privacy, toxicity, fairness, and robustness. To begin with, we apply linear probing to LLMs. The high probing accuracy suggests that LLMs in early pre-training can already distinguish concepts in each trustworthiness dimension. Therefore, to further uncover the hidden possibilities of pre-training, we extract steering vectors from a LLM’s pre-training checkpoints to enhance the LLM’s trustworthiness. Finally, inspired by the theoretical result that mutual information estimation is bounded by linear probing accuracy, we also probe LLMs with mutual information to investigate the dynamics of trustworthiness during pre-training. We are the first to observe a similar two-phase phenomenon: fitting and compression. This research provides an initial exploration of trustworthiness modeling during LLM pre-training, seeking to unveil new insights and spur further developments in the field.