20 results for “steered-teacher”
CS papers onlyHybrid search: Keyword + semantic, ranked by combined score.ⓘ
Want pure semantic search? Try claim verification →
This paper proposes Steered-Teacher On-policy Self-Distillation (ST-OPSD) to address failure modes in on-policy self-distillation.
The paper demonstrates that subliminal learning, where a student model acquires a teacher's traits from semantically unrelated outputs, is fundamentally mediated by a single, transferable steering vec…
Develops a system with subject-aware prompting for adaptive education, improving instructional efficiency and pedagogical quality.
Haolei Xu, Xiaowen Xu, Haiwen Hong, Zixuan Ni +4 more
The paper introduces Relay On-Policy Distillation (Relay-OPD), a method for on-policy distillation that constructs relay trajectories to address prefix failure and improve performance.
This paper proposes Policy Gradient Steering (PGS), a new method for dynamically changing a large language model's behavior at inference time using reinforcement learning.
Yanze Zhang, Hao Liu, Haochang Hao, Zhixiang Shen +2 more
The paper presents a solution to the MetaFine Open Competition's fine-grained manipulation tasks using a scripted teacher and imitation learning.
This paper proposes a source-free continual test-time adaptation methodology using a controlled teacher adaptation and class prototype estimation.
Lan Luo, Anqi Wang, Muzhi Zhou, Junhua Zhu +3 more
This paper presents a system called ParaTutor that supports parent-child tutoring with LLMs, preserving the distinct roles of parents and children.
This paper introduces instruction-based vector steering to redistribute attention in Large Audio-Language Models (LALMs) and demonstrates its effectiveness in recovering sound event locations.
Yeon Su Park, Sieun Kim, Keighley Overbay, Seoyoung Kim +3 more
The paper explores the use of AI-powered automated feedback on online tutoring platforms and its impact on tutors.
SkillC introduces a Contrastive Skill Credit Assignment (CSCA) framework to enable LLM agents to autonomously internalize skills during training, significantly outperforming existing methods without r…
Kun Liang, Chenming Tang, Clive Bai, Weijie Liu +2 more
ADWIN introduces an adaptive window framework for on-policy distillation (OPD) that efficiently manages the supervision horizon by training on short, teacher-anchored prefixes while using delayed full…
The paper proposes CTRL-STEER, a closed-loop framework that adaptively adjusts intervention strength to stabilize concept regulation and improve task success in Vision-Language-Action models without r…
Meng Chen, Anya Ji, Tsung-Han Wu, Tobias Maringgele +3 more
The paper introduces DigitalCoach, a dataset of human expert-novice computer use coaching sessions, and evaluates the ability of state-of-the-art models to teach humans how to use computers.
This paper introduces skill-space shooting, a method that uses foundation model guidance to explore corrections through reusable skills and improve policies autonomously.
Weixian Xu, Yanzhe Zhang, Zora Zhiruo Wang, Changyu Chen +1 more
This paper introduces Sherpa, a multi-turn reinforcement learning framework that trains large language models to adapt their instruction to diverse students and become better aligned with human teache…
Canran Wang, Yuwen Yang, Zhen Wang, Ming Ma +4 more
The paper designs and evaluates a triadic LLM-Teacher collaboration system for K-12 writing, finding that strategic labor division between the LLM and teacher effectively improves writing quality but…
Sina Mavali, David Pape, Jonathan Evertz, Samira Abedini +4 more
The paper introduces the Task Alignment Benchmark (TAB) to evaluate terminal agents' ability to selectively follow relevant environmental instructions while ignoring misleading distractors, revealing…
Jaward Sesay, Yue Yu, Siwei Dong, Yemin Shi +2 more
This paper proposes LectūraAgents, a multi-agent framework for personalized learning through adaptive embodied teaching.