← Back to all news
arXiv Computation and Language

Don't Guess, Ask Wisely: Benchmarking and Training Proactive Elicitation in Open-Domain LLM Conversations

The paper introduces VagueBench, a benchmark that degrades seed queries into vague variants to evaluate large language models’ ability to ask follow‑up questions when user queries are incomplete. It proposes DialogGRPO, an offline reinforcement learning approach that models multi‑turn elicitation as a dialogue‑state directed acyclic graph, using dense turn‑level rewards to guide question selection. DialogGRPO achieves over 88% success in eliciting missing information on vague queries, improves answer coverage and recovery efficiency, and outperforms prompting and test‑time scaling baselines while generalizing zero‑shot to other dialogue datasets.

Related stories

arXiv Machine Learning
Oct 23

GFlowPO: Generative Flow Network as a Language Model Prompt Optimizer

GFlowPO introduces a probabilistic framework for optimizing language model prompts by treating prompt search as posterior inference over latent prompts, regularized by a meta-prompted reference-LM prior. The method first fine‑tunes a lightweight prompt‑LM using an off‑policy Generative Flow Network objective that reuses past prompt evaluations for sample‑efficient exploration. It then applies a training‑free Dynamic Memory Update to refine the meta‑prompt with diverse and top‑performing prompts, leading to consistent improvements over recent discrete prompt optimization baselines on few‑shot classification, instruction induction, and question‑answering tasks.

By Junmo Cho, Suhan Kim, Sangjune An, Minsu Kim, Dong Bok Lee, Heejun Lee, Sung Ju Hwang, Hae Beom Lee
llmsnlpreinforcement-learningfine-tuningbenchmarks
More like this →
arXiv AI
Oct 1

StateTree: Enhancing Long-Term Dialogue Reasoning via Reinforcement Learning

StateTree is a reinforcement learning approach that improves long‑term dialogue reasoning by building a tree‑structured auxiliary task from limited dialogue data. The method embeds key‑value records across multiple sessions into a binary tree, requiring the model to traverse from root to leaf, retrieve records, compare timestamps, and identify a target question among distractors. Curriculum RL training increases tree depth, and a compositional variant trains the model to combine partial reasoning fragments, enabling cross‑session retrieval, temporal reasoning, knowledge updates, and multi‑hop reasoning while generalizing from 10K‑token to 128K‑token contexts.

By Naen Xu, Wanqing Cui, Yibo Hu, Shixin Hong, Hengyu An, Meiguang Jin, Junfeng Ma, Tianyu Du
llmsreinforcement-learningfine-tuning
More like this →
arXiv AI
Oct 23

A Small Model Asks Better: Fine-Tuning Query Clarification for Retrieval-Augmented Generation

The paper introduces ClariEdge, a dataset of about 3,900 examples that capture ambiguous or ill‑formed user queries in retrieval‑augmented generation contexts, including negative cases that do not need clarification. A lightweight LoRA fine‑tuned model achieves 96.6% classification accuracy, outperforming larger prompted baselines in both classification and clarification generation. The study highlights the importance of hard negatives for distinguishing truly ambiguous queries and demonstrates that the same model can be applied before and after retrieval in a RAG pipeline.

By Lihong He, Lucian Popa, Marina Danilevsky
ragfine-tuning
More like this →