8 papers
Does On-Policy Distillation Need to Complete Every Rollout?
Zhengxin Zhang, Maosen Tang, Ning Wang +6
On-policy distillation (OPD) transfers teacher knowledge to a student through feedback on student-generated rollouts, but generating complete rollouts and computing per-token teach…
HARP: Efficient Data Selection for Finetuning Large Language Models
Ning Wang, Zhengxin Zhang, Maosen Tang +3
Finetuning data selection requires balancing two competing goals: selecting examples that improve the downstream objective, and doing so without repeatedly finetuning models. Train…
How Far Are We From True Auto-Research?
Zhengxin Zhang, Ning Wang, Sainyam Galhotra +1
Recent auto-research systems can produce complete papers, but feasibility is not the same as quality, and the field still lacks a systematic study of how good agent-generated paper…
AutoPrunedRetriever: Persistent Reasoning Graphs for Efficient and Extensible Retrieval-Augmented Generation
Ning Wang, Kuanyan Zhu, Daniel Yuehwoon Yee +4
Retrieval-augmented generation (RAG) has become a standard approach for knowledge-intensive LLM tasks. Complex reasoning remains challenging because queries often require composing…
TableGPT-R1: Advancing Tabular Reasoning Through Reinforcement Learning
Saisai Yang, Qingyi Huang, Jing Yuan +13
Tabular data serves as the backbone of modern data analysis and scientific research. While Large Language Models (LLMs) fine-tuned via Supervised Fine-Tuning (SFT) have significant…
MetaJoin: Adaptive Join Selections for Heterogeneous Tables
Ning Wang, Sainyam Galhotra
Joining heterogeneous tables requires different forms of correspondence, including exact agreement, syntactic transformation, semantic similarity, and entity-level matching. The ap…
REAP: Enhancing RAG with Recursive Evaluation and Adaptive Planning for Multi-Hop Question Answering
Yijie Zhu, Haojie Zhou, Wanting Hong +2
Retrieval-augmented generation (RAG) has been extensively employed to mitigate hallucinations in large language models (LLMs). However, existing methods for multi-hop reasoning tas…
Kongzi: A Historical Large Language Model with Fact Enhancement
Jiashu Yang, Ningning Wang, Yian Zhao +6
Large Language Models (LLMs) can exhibit factual inaccuracies during long-chain reasoning or under misleading prompts, limiting their reliability in historical analysis. We propose…