More Like this

1. GuideLLM: Exploring LLM-Guided Conversation with Applications in Autobiography Interviewing

   Duan, J; Zhao, X; Zhang, Z; Ko, E; Boddy, L; Wang, C; Li, T; Rasgon, A; Hong, J; Lee, M K; et al (February 2025, https://doi.org/10.48550/arXiv.2502.06494)

   Although Large Language Models (LLMs) succeed in human-guided conversations such as instruction following and question answering, the potential of LLM-guided conversations-where LLMs direct the discourse and steer the conversation's objectives-remains under-explored. In this study, we first characterize LLM-guided conversation into three fundamental components: (i) Goal Navigation; (ii) Context Management; (iii) Empathetic Engagement, and propose GuideLLM as an installation. We then implement an interviewing environment for the evaluation of LLM-guided conversation. Specifically, various topics are involved in this environment for comprehensive interviewing evaluation, resulting in around 1.4k turns of utterances, 184k tokens, and over 200 events mentioned during the interviewing for each chatbot evaluation. We compare GuideLLM with 6 state-of-the-art LLMs such as GPT-4o and Llama-3-70b-Instruct, from the perspective of interviewing quality, and autobiography generation quality. For automatic evaluation, we derive user proxies from multiple autobiographies and employ LLM-as-a-judge to score LLM behaviors. We further conduct a human-involved experiment by employing 45 human participants to chat with GuideLLM and baselines. We then collect human feedback, preferences, and ratings regarding the qualities of conversation and autobiography. Experimental results indicate that GuideLLM significantly outperforms baseline LLMs in automatic evaluation and achieves consistent leading performances in human ratings. 

   more » « less
2. Magpie: Generating High-Quality Synthetic Data With Open-Source Large Language Models

   Xu, Zhangchen (June 2025, University of Washington)

   High-quality instruction data is critical for aligning large language models (LLMs). Although some models, such as Llama-3-Instruct, have open weights, their alignment data remain private, which hinders the democratization of AI. High human labor costs and a limited, predefined scope for prompting prevent existing open-source data creation methods from scaling effectively, potentially limiting the diversity and quality of public alignment datasets. Is it possible to synthesize high-quality instruction data at scale by extracting it directly from an aligned LLM? We present a self-synthesis method for generating large-scale alignment data named Magpie. Our key observation is that aligned LLMs like Llama-3-Instruct can generate a user query when we input only the pre-query templates up to the position reserved for user messages, thanks to their auto-regressive nature. We use this method to prompt Llama-3-Instruct and generate 4 million instructions along with their corresponding responses. We further introduce extensions of Magpie for filtering, generating multi-turn, preference optimization, domain-specific and multilingual datasets. We perform a comprehensive analysis of the Magpie-generated data. To compare Magpie-generated data with other public instruction datasets (eg, ShareGPT, WildChat, Evol-Instruct, Ultra-Chat, OpenHermes, Tulu-V2-Mix, GenQA), we fine-tune Llama-3-8B-Base with each dataset and evaluate the performance of the fine-tuned models. Our results indicate that using Magpie for supervised fine-tuning (SFT) solely can surpass the performance of previous public datasets utilized for both SFT and preference optimization, such as direct preference optimization with UltraFeedback. We also show that in some tasks, models supervised fine-tuned with Magpie perform comparably to the official Llama-3-8B-Instruct, despite the latter being enhanced with 10 million data points through SFT and subsequent preference optimization. This advantage is evident on alignment benchmarks such as AlpacaEval, ArenaHard, and WildBench. 

   more » « less
3. Improving Data Efficiency via Curating LLM-Driven Rating Systems

   Pang, Jinlong; Wei, Jiaheng; Shah, Ankit; Zhu, Zhaowei; Wang, Yaxuan; Qian, Chen; Liu, Yang; Bao, Yujia; Wei, Wei (April 2025, The Thirteenth International Conference on Learning Representations)

   Instruction tuning is critical for adapting large language models (LLMs) to downstream tasks, and recent studies have demonstrated that small amounts of human-curated data can outperform larger datasets, challenging traditional data scaling laws. While LLM-based data quality rating systems offer a cost-effective alternative to human annotation, they often suffer from inaccuracies and biases, even in powerful models like GPT-4. In this work, we introduce DS2, a Diversity-aware Score curation method for Data Selection. By systematically modeling error patterns through a score transition matrix, DS2 corrects LLM-based scores and promotes diversity in the selected data samples. Our approach shows that a curated subset (just 3.3% of the original dataset) outperforms full-scale datasets (300k samples) across various machine-alignment benchmarks, and matches or surpasses human-aligned datasets such as LIMA with the same sample size (1k samples). These findings challenge conventional data scaling assumptions, highlighting that redundant, low-quality samples can degrade performance and reaffirming that "more can be less." 

   more » « less
4. Demo: An Exploration of LLM-Guided Conversation in Reminiscence Therapy

   Duan, J; Zhao, X; Zhang, Z; Ko, E; Boddy, L; Wang, C; Li, T; Rasgon, A; Hong, J; Lee, M K; et al (December 2024, Gen AI f or Health Workshop @ NeurIPS 2024, Vancouver.)

   Although Large Language Models (LLMs) succeed in human-guided conversations such as instruction following and question answering, the potential of LLM-guided conversations—where LLMs direct the discourse and steer the conversation’s objectives—remains largely untapped. In this study, we provide an exploration of the LLM-guided conversation paradigm. Specifically, we first characterize LLM-guided conversation into three fundamental properties: (i) Goal Navigation; (ii) Context Management; (iii) Empathetic Engagement, and propose GUIDELLM as a general framework for LLM-guided conversation. We then implement an autobiography interviewing environment as one of the demonstrations of GuideLLM, which is a common practice in Reminiscence Therapy. In this environment, various techniques are integrated with GUIDELLM to enhance the autonomy of LLMs, such as Verbalized Interview Protocol (VIP) and Memory Graph Extrapolation (MGE) for goal navigation, and therapy strategies for empathetic engagement. We compare GUIDELLM with baseline LLMs, such as GPT-4-turbo and GPT-4o, from the perspective of interviewing quality, conversation quality, and autobiography generation quality. Experimental results encompassing both LLM-as-a-judge evaluations and human subject experiments involving 45 participants indicate that GUIDELLM significantly outperforms baseline LLMs in the autobiography interviewing task. 

   more » « less
5. SemCoder: Training Code Language Models with Comprehensive Semantics Reasoning

   Ding, Yangruibo; Peng, Jinjun; Min, Marcus J; Kaiser, Gail; Yang, Junfeng; Ray, Baishakhi (September 2024, OpenReview.net)

   Code Large Language Models (Code LLMs) have excelled at tasks like code completion but often miss deeper semantics such as execution effects and dynamic states. This paper aims to bridge the gap between Code LLMs' reliance on static text data and the need for semantic understanding for complex tasks like debugging and program repair. We introduce a novel strategy, monologue reasoning, to train Code LLMs to reason comprehensive semantics, encompassing high-level functional descriptions, local execution effects of individual statements, and overall input/output behavior, thereby linking static code text with dynamic execution states. We begin by collecting PyX, a clean Python corpus of fully executable code samples with functional descriptions and test cases. We propose training Code LLMs not only to write code but also to understand code semantics by reasoning about key properties, constraints, and execution behaviors using natural language, mimicking human verbal debugging, i.e., rubber-duck debugging. This approach led to the development of SemCoder, a Code LLM with only 6.7B parameters, which shows competitive performance with GPT-3.5-turbo on code generation and execution reasoning tasks. SemCoder achieves 79.3% on HumanEval (GPT-3.5-turbo: 76.8%), 63.6% on CRUXEval-I (GPT-3.5-turbo: 50.3%), and 63.9% on CRUXEval-O (GPT-3.5-turbo: 59.0%). We also study the effectiveness of SemCoder's monologue-style execution reasoning compared to concrete scratchpad reasoning, showing that our approach integrates semantics from multiple dimensions more smoothly. Finally, we demonstrate the potential of applying learned semantics to improve Code LLMs' debugging and self-refining capabilities. Our data, code, and models are available at: https://github.com/ARiSE-Lab/SemCoder. 

   more » « less