Guanhua Chen

Ying Zhao, Guanhua Chen, Jie Liu

The pursuit of advanced polymers for energy technologies, spanning photovoltaics, solid-state batteries, and hydrogen storage, is hindered by fragmented data ecosystems that fail to capture the hierarchical complexity of these materials. Polymer science lacks interoperable databases, forcing reliance on disconnected literature and legacy records riddled with unstructured formats and irreproducible testing protocols. This fragmentation stifles machine learning (ML) applications and delays the discovery of materials critical for global decarbonization. Three systemic barriers compound the challenge. First, academic-industrial data silos restrict access to proprietary industrial datasets, while academic publications often omit critical synthesis details. Second, inconsistent testing methods undermine cross-study comparability. Third, incomplete metadata in existing databases limits their utility for training reliable ML models. Emerging solutions address these gaps through technological and collaborative innovation. Natural language processing (NLP) tools extract structured polymer data from decades of literature, while high-throughput robotic platforms generate self-consistent datasets via autonomous experimentation. Central to these advances is the adoption of FAIR (Findable, Accessible, Interoperable, Reusable) principles, adapted to polymer-specific ontologies, ensuring machine-readability and reproducibility. Future breakthroughs hinge on cultural shifts toward open science, accelerated by decentralized data markets and autonomous laboratories that merge robotic experimentation with real-time ML validation. By addressing data fragmentation through technological innovation, collaborative governance, and ethical stewardship, the polymer community can transform bottlenecks into accelerants.

Zeguan Xiao, Yun Chen, Guanhua Chen et al.

Direct Alignment Algorithms (DAAs), such as Direct Preference Optimization (DPO) and Simple Preference Optimization (SimPO), have emerged as efficient alternatives to Reinforcement Learning from Human Feedback (RLHF) algorithms for aligning large language models (LLMs) with human preferences. However, DAAs suffer from a fundamental limitation we identify as the "reward-generation gap" -- a misalignment between optimization objectives during training and actual generation performance during inference. In this paper, we find a contributor to the reward-generation gap is the mismatch between the inherent importance of prefix tokens during the LLM generation process and how this importance is reflected in the implicit reward functions of DAAs. To bridge the gap, we adopt a token-level MDP perspective of DAAs to analyze its limitations and introduce a simple yet effective approach called Prefix-Oriented Equal-length Training (POET), which truncates both preferred and dispreferred responses to match the shorter one's length. Training with \mname, where both responses in each sample are truncated to equal length, resulting in diverse truncated lengths across samples, the optimization of DAAs objective is implicitly constrained to converge across all timesteps of token-level MDP, thus paying more attention to prefix tokens than the standard DAAs. We conduct experiments with DPO and SimPO, two representative DAAs, demonstrating that POET improves over their standard implementations, achieving up to 15.6 points in AlpacaEval 2 and overall improvements across downstream tasks. Our results highlight the importance of addressing the misalignment between reward optimization and generation performance in DAAs.