Vertical-domain few-shot classification remains challenging for small language models, as limited supervision makes it difficult to acquire domain-specific decision knowledge. On-Policy Distillation (OPD) can improve teacher-guided adaptation by supervising student-generated rollouts, while GRPO-based reinforcement learning can further refine downstream predictions. However, existing KD-to-RL pipelines typically rely on globally fixed transition schedules, ignoring that different samples may require different amounts of teacher-guided acquisition before reward-driven refinement. We propose PIVOT (Perplexity-Informed Transition Optimization), a dynamic transition framework that routes samples between OPD and GRPO according to teacher-evaluated sequence perplexity. PIVOT moves low-perplexity samples to GRPO for reward-driven refinement while keeping high-perplexity samples under OPD for continued domain knowledge acquisition. Experiments on Banking77 and HWU64 show that PIVOT consistently outperforms continued OPD and globally synchronized OPD→GRPO baselines under the same optimization budget, achieving stronger downstream performance and more stable training dynamics.

Despite recent advances in safety alignment, large language models (LLMs) remain highly susceptible to adversarial attacks, while the internal mechanisms behind such vulnerabilities are still poorly understood. Existing gradient-based attribution methods offer valuable interpretability for analyzing information storage and processing in LLMs. However, they are inapplicable to adversarial attacks, which typically occur in open-ended generation settings without fixed ground-truth outputs. To address these challenges, we propose a novel similarity-based gradient attribution method to identify key neurons sensitive to adversarial behaviors in open-ended generation tasks. The detected neurons, termed targeted neurons, play a critical role in safety training. Building on this neuron-level perspective, we uncover two key neuronal patterns: (i) universal neurons that are consistently exploited across multiple attack strategies, and (ii) interference neurons that hinder safety improvements when fine-tuned indiscriminately, providing mechanistic insights into the interpretability of adversarial vulnerabilities. Inspired by these findings, we propose a neuron-level defense strategy, Targeted Neuron Tuning (TNT), which selectively fine-tunes the identified targeted neurons for specific attacks. Experimental evaluations across multiple LLM architectures and scales demonstrate that TNT substantially improves model robustness against a wide range of jailbreak attacks, achieving safe rates exceeding 90% and even approaching 100%, while preserving general task performance, enabling precise and robust safety interventions. Warning: This paper contains example data that may be harmful.

Voice agents typically use text-first pipeline for safety, but large language model (LLM) outputs are often overly written, making even high-quality text-to-speech (TTS) sound robotic. This is especially problematic in domains such as healthcare and sales, where different audiences require graded orality control. Text-level rewriting can control spoken style, but resources mostly study the reverse direction (spoken-to-written transfer), and its internal locus in LLMs remains unclear. We study English written-to-spoken rewriting with graded orality control. We introduce WSPairs, 4,709 written–spoken pairs from spontaneous conversations, annotated with 0–5 Orality Levels using seven sociolinguistic features. Through causal activation patching and head-level attribution, we find that orality concentrates in late layers and a sparse set of high-influence attention heads. We then propose CAS, a no-weight-update calibrated activation-steering method, and LAS, a learned activation-steering method. CAS improves exact level targeting and reduces bias relative to few-shot prompting, although target ordering remains imperfect at the highest levels. LAS uses 0.004%–0.006% trainable parameters, obtains higher Orality Level Accuracy (L-Acc) than full SFT on Qwen-3B and Mistral-7B, remains the strongest lightweight method on Llama-1B, and leads SFT by up to 3.9 L-Acc points out of domain. These results support localized, efficient control of text-level spoken style. Dataset and code are available.

Currently, most agentic systems are driven by mainstream closed-source large language models (LLMs), such as GPT-5 and Gemini 3, resulting in high operational costs. As a result, deploying customized small language models (SLMs) to replace LLMs in agent components has attracted increasing attention. However, the limited capabilities of SLMs often render them inadequate for agent tasks. To address this, we propose Edit4Agency, a novel framework that utilizes model editing to efficiently and effectively enhance agent sub-capabilities of SLMs. By formalizing tasks as a sequence generation problem triggered by specific conditions, the framework embeds task triggers into system prompts and performs model editing through localized parameter updates, eliminating the need for large-scale retraining. Extensive experiments across multiple datasets and case studies in agent systems demonstrate that model editing, with only 50 data samples, enables a base model without instruction tuning to acquire new skills such as tool calling while maintaining its general capabilities. The results show that Edit4Agency outperforms instruction-tuned SLMs, constrained decoding, and SLMs trained with large-scale supervised fine-tuning or reinforcement learning methods.

Dialogue agents powered by Large Language Models (LLMs) show superior performance in various tasks. Despite the better user understanding and human-like responses, their lack of controllability remains a key challenge, often leading to unfocused conversations or task failure. To address this, we introduce Standard Operating Procedure (SOP) to regulate dialogue flow. Specifically, we propose ChatSOP, a novel SOP-guided Monte Carlo Tree Search (MCTS) planning framework designed to enhance the controllability of LLM-driven dialogue agents. To enable this, we curate a dataset comprising SOP-annotated multi-scenario dialogues, generated using a semi-automated role-playing system with GPT-4o and validated through strict manual quality control. Additionally, we propose a novel method that integrates Chain of Thought reasoning with supervised fine-tuning for SOP prediction and utilizes SOP-guided Monte Carlo Tree Search for optimal action planning during dialogues. Experimental results demonstrate the effectiveness of our method, such as achieving a 27.95% improvement in action accuracy compared to baseline models based on GPT-3.5 and also showing notable gains for open-source models. Dataset and codes are publicly available.

Paraphrase generation is a longstanding NLP task and achieves great success with the aid of large corpora. However, transferring a paraphrasing model to another domain encounters the problem of domain shifting especially when the data is sparse. At the same time, widely using large pre-trained language models (PLMs) faces the overfitting problem when training on scarce labeled data. To mitigate these two issues, we propose, LAPA, an effective adapter for PLMs optimized by meta-learning. LAPA has three-stage training on three types of related resources to solve this problem: 1. pre-training PLMs on unsupervised corpora, 2. inserting an adapter layer and meta-training on source domain labeled data, and 3. fine-tuning adapters on a small amount of target domain labeled data. This method enables paraphrase generation models to learn basic language knowledge first, then learn the paraphrasing task itself later, and finally adapt to the target task. Our experimental results demonstrate that LAPA achieves state-of-the-art in supervised, unsupervised, and low-resource settings on three benchmark datasets. With only 2% of trainable parameters and 1% labeled data of the target task, our approach can achieve a competitive performance with previous work.