Total relevant papers: 3
Paper selection prompt and criteria at the bottom
Table of contents with paper titles:
D2PO: Discriminator-Guided DPO with Response Evaluation Models Authors: Prasann Singhal, Nathan Lambert, Scott Niekum, Tanya Goyal, Greg Durrett
FLAME: Factuality-Aware Alignment for Large Language Models Authors: Sheng-Chieh Lin, Luyu Gao, Barlas Oguz, Wenhan Xiong, Jimmy Lin, Wen-tau Yih, Xilun Chen
SHED: Shapley-Based Automated Dataset Refinement for Instruction Fine-Tuning Authors: Yexiao He, Ziyao Wang, Zheyu Shen, Guoheng Sun, Yucong Dai, Yongkai Wu, Hongyi Wang, Ang Li
ArXiv ID: 2405.01511 Authors: Prasann Singhal, Nathan Lambert, Scott Niekum, Tanya Goyal, Greg Durrett
Abstract: arXiv:2405.01511v1 Announce Type: new Abstract: Varied approaches for aligning language models have been proposed, including supervised fine-tuning, RLHF, and direct optimization methods such as DPO. Although DPO has rapidly gained popularity due to its straightforward training process and competitive results, there is an open question of whether there remain practical advantages of using a discriminator, like a reward model, to evaluate responses. We propose D2PO, discriminator-guided DPO, an approach for the online setting where preferences are being collected throughout learning. As we collect gold preferences, we use these not only to train our policy, but to train a discriminative response evaluation model to silver-label even more synthetic data for policy training. We explore this approach across a set of diverse tasks, including a realistic chat setting, we find that our approach leads to higher-quality outputs compared to DPO with the same data budget, and greater efficiency in terms of preference data requirements. Furthermore, we show conditions under which silver labeling is most helpful: it is most effective when training the policy with DPO, outperforming traditional PPO, and benefits from maintaining a separate discriminator from the policy model.
Comment: This paper is relevant to criterion 1 as it discusses a new methodological improvement to RLHF, specifically a discriminator-guided DPO approach for aligning language models. Relevance: 8 Novelty: 6
ArXiv ID: 2405.01525 Authors: Sheng-Chieh Lin, Luyu Gao, Barlas Oguz, Wenhan Xiong, Jimmy Lin, Wen-tau Yih, Xilun Chen
Abstract: arXiv:2405.01525v1 Announce Type: new Abstract: Alignment is a standard procedure to fine-tune pre-trained large language models (LLMs) to follow natural language instructions and serve as helpful AI assistants. We have observed, however, that the conventional alignment process fails to enhance the factual accuracy of LLMs, and often leads to the generation of more false facts (i.e. hallucination). In this paper, we study how to make the LLM alignment process more factual, by first identifying factors that lead to hallucination in both alignment steps:\ supervised fine-tuning (SFT) and reinforcement learning (RL). In particular, we find that training the LLM on new knowledge or unfamiliar texts can encourage hallucination. This makes SFT less factual as it trains on human labeled data that may be novel to the LLM. Furthermore, reward functions used in standard RL can also encourage hallucination, because it guides the LLM to provide more helpful responses on a diverse set of instructions, often preferring longer and more detailed responses. Based on these observations, we propose factuality-aware alignment, comprised of factuality-aware SFT and factuality-aware RL through direct preference optimization. Experiments show that our proposed factuality-aware alignment guides LLMs to output more factual responses while maintaining instruction-following capability.
Comment: Matches criterion 1 closely as it discusses factuality-aware alignment for instruction-following in LLMs. Relevance: 8 Novelty: 6
ArXiv ID: 2405.00705 Authors: Yexiao He, Ziyao Wang, Zheyu Shen, Guoheng Sun, Yucong Dai, Yongkai Wu, Hongyi Wang, Ang Li
Abstract: arXiv:2405.00705v1 Announce Type: new Abstract: The pre-trained Large Language Models (LLMs) can be adapted for many downstream tasks and tailored to align with human preferences through fine-tuning. Recent studies have discovered that LLMs can achieve desirable performance with only a small amount of high-quality data, suggesting that a large amount of the data in these extensive datasets is redundant or even harmful. Identifying high-quality data from vast datasets to curate small yet effective datasets has emerged as a critical challenge. In this paper, we introduce SHED, an automated dataset refinement framework based on Shapley value for instruction fine-tuning. SHED eliminates the need for human intervention or the use of commercial LLMs. Moreover, the datasets curated through SHED exhibit transferability, indicating they can be reused across different LLMs with consistently high performance. We conduct extensive experiments to evaluate the datasets curated by SHED. The results demonstrate SHED's superiority over state-of-the-art methods across various tasks and LLMs; notably, datasets comprising only 10% of the original data selected by SHED achieve performance comparable to or surpassing that of the full datasets.
Comment: Matches criterion 1 closely as it discusses a method for instruction fine-tuning. Relevance: 8 Novelty: 6
In suggesting papers to your friend, remember that he enjoys papers on statistical machine learning, and generative modeling in natural language processing. Your friend also likes learning about surprising empirical results in language models, as well as clever statistical tricks. He does not want to read papers that are about primarily applications of methods to specific domains.