Recently, several multi-turn dialogue benchmarks have been proposed to evaluate the conversational abilities of large language models (LLMs). As LLMs are increasingly recognized as key technology for advancing intelligent education, owing to their ability to deeply understand instructional contexts and provide personalized guidance, constructing dedicated teacher-student dialogue benchmarks has become particularly important. To this end, we present EduDial, a comprehensive multi-turn teacher-student dialogue dataset. EduDial covers 345 core knowledge points and consists of 34,250 dialogue sessions from interactions between teacher and student agents. Its design is guided by Bloom’s taxonomy of educational objectives and incorporates ten questioning strategies—including situational questioning, zone of proximal development (ZPD) questioning, and metacognitive questioning—better capturing authentic classroom interactions. Furthermore, we design differentiated teaching strategies for students at different cognitive levels, providing more targeted guidance. Building on EduDial, we develop EduDial-LLM 32B via training and propose an 11-dimensional evaluation framework that systematically measures the teaching abilities of LLMs, encompassing overall teaching quality and content quality. Experiments on 17 mainstream LLMs reveal that most models struggle in student-centered teaching scenarios, whereas EduDial-LLM achieves significant gains, consistently outperforming all baselines across all metrics. Code is available at https://github.com/ECNU-RAIL/EduDial-EMNLP2026.

Large language models (LLMs) are shifting from answer providers to intelligent tutors in educational settings. However, current supervised fine-tuning methods only learn surface-level teaching patterns without dynamic adaptation. Recent reinforcement learning approaches address this, yet two key challenges remain. First, they evaluate teaching effectiveness solely based on whether students produce correct outputs, unable to distinguish whether students genuinely understand or echo teacher-provided answers. Second, they cannot perceive students’ evolving cognitive states through dialogue, failing to adapt teaching strategies accordingly. We propose Unidirectional Cognitive Optimization (UCO) to address these challenges. UCO employs a multi-turn interactive reinforcement learning paradigm with two synergistic reward functions: the Progress Reward captures cognitive advancement, evaluating students’ transition from confusion to comprehension, and the Scaffold Reward dynamically identifies each student’s Zone of Proximal Development (ZPD), encouraging the teacher to conduct productive teaching within this zone. We evaluate UCO against 11 baseline models on BigMath and MathTutorBench. Results demonstrate that UCO outperforms all equivalent-scale models and achieves performance comparable to leading closed-source models. Code is available at https://github.com/ECNU-RAIL/UCO-EMNLP2026.

Sentence representation learning is a fundamental task in NLP. Existing methods use contrastive learning (CL) to learn effective sentence representations, which benefit from high-quality contrastive data but require extensive human annotation. Large language models (LLMs) like ChatGPT and GPT4 can automatically generate such data. However, this alternative strategy also encounters challenges: 1) obtaining high-quality generated data from small-parameter LLMs is difficult, and 2) inefficient utilization of the generated data. To address these challenges, we propose a novel adaptive reinforcement tuning (ART) framework. Specifically, to address the first challenge, we introduce a reinforcement learning approach for fine-tuning small-parameter LLMs, enabling the generation of high-quality hard contrastive data without human feedback. To address the second challenge, we propose an adaptive iterative framework to guide the small-parameter LLMs to generate progressively harder samples through multiple iterations, thereby maximizing the utility of generated data. Experiments conducted on seven semantic text similarity tasks demonstrate that the sentence representation models trained using the synthetic data generated by our proposed method achieve state-of-the-art performance. Our code is available at https://github.com/WuNein/AdaptCL.