arXiv subjects

Hongyao Chen

Publications and source records attributed to Hongyao Chen.

4 recordsLinked to original sources

arXiv2026

METER: Evaluating Multi-Level Contextual Causal Reasoning in Large Language Models

Contextual causal reasoning is a critical yet challenging capability for Large Language Models (LLMs). Existing benchmarks, however, often evaluate this skill in fragmented settings, failing to ensure context consistency or cover the full causal hierarchy. To address this, we pioneer METER to systematically benchmark LLMs across all three levels of the causal ladder under a unified context setting. Our extensive evaluation of various LLMs reveals a significant decline in proficiency as tasks ascend the causal hierarchy. To diagnose this degradation, we conduct a deep mechanistic analysis via both error pattern identification and internal information flow tracing. Our analysis reveals two primary failure modes: (1) LLMs are susceptible to distraction by causally irrelevant but factually correct information at lower level of causality; and (2) as tasks ascend the causal hierarchy, faithfulness to the provided context degrades, leading to a reduced performance. We belive our work advances our understanding of the mechanisms behind LLM contextual causal reasoning and establishes a critical foundation for future research. Our code and dataset are available at https://github.com/SCUNLP/METER .

cs.CL↗

arXiv2025

MARS2 2025 Challenge on Multimodal Reasoning: Datasets, Methods, Results, Discussion, and Outlook

This paper reviews the MARS2 2025 Challenge on Multimodal Reasoning. We aim to bring together different approaches in multimodal machine learning and LLMs via a large benchmark. We hope it better allows researchers to follow the state-of-the-art in this very dynamic area. Meanwhile, a growing number of testbeds have boosted the evolution of general-purpose large language models. Thus, this year's MARS2 focuses on real-world and specialized scenarios to broaden the multimodal reasoning applications of MLLMs. Our organizing team released two tailored datasets Lens and AdsQA as test sets, which support general reasoning in 12 daily scenarios and domain-specific reasoning in advertisement videos, respectively. We evaluated 40+ baselines that include both generalist MLLMs and task-specific models, and opened up three competition tracks, i.e., Visual Grounding in Real-world Scenarios (VG-RS), Visual Question Answering with Spatial Awareness (VQA-SA), and Visual Reasoning in Creative Advertisement Videos (VR-Ads). Finally, 76 teams from the renowned academic and industrial institutions have registered and 40+ valid submissions (out of 1200+) have been included in our ranking lists. Our datasets, code sets (40+ baselines and 15+ participants' methods), and rankings are publicly available on the MARS2 workshop website and our GitHub organization page https://github.com/mars2workshop/, where our updates and announcements of upcoming events will be continuously provided.

cs.CV↗

arXiv2025

Hybrid Batch Normalisation: Resolving the Dilemma of Batch Normalisation in Federated Learning

Batch Normalisation (BN) is widely used in conventional deep neural network training to harmonise the input-output distributions for each batch of data. However, federated learning, a distributed learning paradigm, faces the challenge of dealing with non-independent and identically distributed data among the client nodes. Due to the lack of a coherent methodology for updating BN statistical parameters, standard BN degrades the federated learning performance. To this end, it is urgent to explore an alternative normalisation solution for federated learning. In this work, we resolve the dilemma of the BN layer in federated learning by developing a customised normalisation approach, Hybrid Batch Normalisation (HBN). HBN separates the update of statistical parameters (i.e. , means and variances used for evaluation) from that of learnable parameters (i.e. , parameters that require gradient updates), obtaining unbiased estimates of global statistical parameters in distributed scenarios. In contrast with the existing solutions, we emphasise the supportive power of global statistics for federated learning. The HBN layer introduces a learnable hybrid distribution factor, allowing each computing node to adaptively mix the statistical parameters of the current batch with the global statistics. Our HBN can serve as a powerful plugin to advance federated learning performance. It reflects promising merits across a wide range of federated learning settings, especially for small batch sizes and heterogeneous data.

cs.LG↗

arXiv2020

Modern alleles in archaic human Y chromosomes support origin of modern human paternal lineages in Asia rather than Africa

Recent studies have shown that hybridization between modern and archaic humans was commonplace in the history of our species. After admixture, some individuals with admixed autosomes carried the modern Homo Sapiens uniparental DNAs, while the rest carried the archaic versions. Coevolution of admixed autosomes and uniparental DNAs is expected to cause some of the sites in modern uniparental DNAs to revert back to archaic alleles, while the opposite process would occur (from archaic to modern) in some of the sites in archaic uniparental DNAs. This type of coevolution is one of the elements that differentiate the two different models of the Y phylogenetic tree of modern humans, rooting it either in Africa or East Asia. The expected reversion to archaic alleles is assumed to occur and is easily traceable in the Asia model, but is absent in the Africa model due to its infinite site assumption, which also precludes the independent or convergent mutation to modern alleles in archaic uniparental DNAs since mutations are assumed to occur randomly across a neutral genome, and convergent evolution is assumed not to occur. Here, we examined newly published high coverage Y chromosome sequencing data of two Denisovan and two Neanderthal samples to determine whether they carry modern-Homo Sapiens alleles in sites where they are not supposed to according to the Africa model. The results showed that a significant fraction of the sites that, according to the Asia model, should differentiate the original modern Y from the original archaic Y carried modern alleles in the archaic Y samples here. Some of these modern alleles were shared among all archaic humans while others could differentiate Denisovans from Neanderthals. The observation is best accounted for by coevolution of archaic Y and admixed modern autosomes, and hence supports the Asia model, since it takes such coevolution into account.

q-bio.OT↗