arXiv subjects

Xinhua Yang

Publications and source records attributed to Xinhua Yang.

4 recordsLinked to original sources

arXiv2025

ELSPR: Evaluator LLM Training Data Self-Purification on Non-Transitive Preferences via Tournament Graph Reconstruction

Pairwise evaluation of large language models (LLMs) has become the dominant paradigm for benchmarking open-ended tasks, yet non-transitive preferences, where evaluators prefer A over B, B over C, but C over A, fundamentally undermine ranking reliability. We show that this critical issue stems largely from low-quality data that contains inherently ambiguous preference pairs. To address this challenge, we propose ELSPR, a principled graph-theoretic framework that models pairwise preferences as tournament graphs and systematically identifies problematic training data. ELSPR quantifies non-transitivity through strongly connected components (SCCs) analysis and measures overall preference clarity using a novel normalized directed graph structural entropy metric. Our filtering methodology selectively removes preference data that induce non-transitivity while preserving transitive preferences. Extensive experiments on the AlpacaEval benchmark demonstrate that models fine-tuned on ELSPR-filtered data achieve substantial improvements: a 13.8% reduction in non-transitivity, a 0.088 decrease in structural entropy, and significantly enhanced discriminative power in real-world evaluation systems. Human validation confirms that discarded data exhibit dramatically lower inter-annotator agreement (34.4% vs. 52.6%) and model-human consistency (51.2% vs. 80.6%) compared to cleaned data. These findings establish ELSPR as an effective data self-purification approach for developing more robust, consistent, and human-aligned LLM evaluation systems.

cs.CL↗

arXiv2025

MIDB: Multilingual Instruction Data Booster for Enhancing Cultural Equality in Multilingual Instruction Synthesis

Despite doubts on data quality, instruction synthesis has been widely applied into instruction tuning (IT) of LLMs as an economic and rapid alternative. Recent endeavors focus on improving data quality for synthesized instruction pairs in English and have facilitated IT of English-centric LLMs. However, data quality issues in multilingual synthesized instruction pairs are even more severe, since the common synthesizing practice is to translate English synthesized data into other languages using machine translation (MT). Besides the known content errors in these English synthesized data, multilingual synthesized instruction data are further exposed to defects introduced by MT and face insufficient localization of the target languages, leading to cultural inequality in trained LLMs. In this paper, we propose MIDB, a Multilingual Instruction Data Booster to automatically address the quality issues in multilingual synthesized data. MIDB is trained on around 36.8k revision examples across 16 languages by human linguistic experts, thereby can boost the low-quality data by addressing content errors and MT defects, and improving localization in these synthesized data. Both automatic and human evaluation indicate that not only MIDB steadily improved instruction data quality in 16 languages, but also the instruction-following and cultural-understanding abilities of multilingual LLMs fine-tuned on MIDB-boosted data were significantly enhanced, suggesting an improved linguistic and cultural equality.

cs.CL↗

arXiv2025

Adapting Large Language Models to Log Analysis with Interpretable Domain Knowledge

Log analysis represents a critical sub-domain within AI applications that facilitates automatic approaches to fault and error management of large-scaled software systems, saving labors of traditional manual methods. While existing solutions using large language models (LLMs) show promise, they are limited by a significant domain gap between natural and log languages (the latter contains rich domain-specific tokens such as status codes, IP addresses, resource pathes), which restricts their effectiveness in real-world applications. However, directly adapting general-purpose LLMs to log analysis using raw logs may degrade their performance due to inconsistent token distribution. In this paper, we present a domain adaptation approach that addresses these limitations by integrating interpretable domain knowledge into open-source LLMs through continual pre-training (CPT), which bridges this domain gap by adapting LLMs on interpretable natural texts with log knowledge (instead of raw logs) to reduce distribution discrepancy. To achieve this, we developed NLPLog, a comprehensive dataset containing over 250,000 question-answer pairs on log-related knowledge. Our resulting model, SuperLog, achieves the best performance across four log analysis tasks, with an average accuracy improvement of 12.01% over the second-best model. Ablation study also suggests advantages of domain adaption using interpretable log knowledge over using raw logs.

cs.CL↗

arXiv2015

Exploring genetic variation in the tomato (Solanum section Lycopersicon) clade by whole-genome sequencing

Genetic variation in the tomato clade was explored by sequencing a selection of 84 tomato accessions and related wild species representative for the Lycopersicon, Arcanum, Eriopersicon, and Neolycopersicon groups. We present a reconstruction of three new reference genomes in support of our comparative genome analyses. Sequence diversity in commercial breeding lines appears extremely low, indicating the dramatic genetic erosion of crop tomatoes. This is reflected by the SNP count in wild species which can exceed 10 million i.e. 20 fold higher than in crop accessions. Comparative sequence alignment reveals group, species, and accession specific polymorphisms, which explain characteristic fruit traits and growth habits in tomato accessions. Using gene models from the annotated Heinz reference genome, we observe a bias in dN/dS ratio in fruit and growth diversification genes compared to a random set of genes, which probably is the result of a positive selection. We detected highly divergent segments in wild S. lycopersicum species, and footprints of introgressions in crop accessions originating from a common donor accession. Phylogenetic relationships of fruit diversification and growth specific genes from crop accessions show incomplete resolution and are dependent on the introgression donor. In contrast, whole genome SNP information has sufficient power to resolve the phylogenetic placement of each accession in the four main groups in the Lycopersicon clade using Maximum Likelihood analyses. Phylogenetic relationships appear correlated with habitat and mating type and point to the occurrence of geographical races within these groups and thus are of practical importance for introgressive hybridization breeding. Our study illustrates the need for multiple reference genomes in support of tomato comparative genomics and Solanum genome evolution studies.

q-bio.GN↗