arXiv subjects

Nan He

Publications and source records attributed to Nan He.

7 recordsLinked to original sources

arXiv2026

Test-time Scaling over Perception: Resolving the Grounding Paradox in Thinking with Images

Recent multimodal large language models (MLLMs) support Thinking with Images, invoking visual tools such as zooming and cropping to inspect image regions during inference. Yet these systems remain brittle in fine-grained reasoning: to acquire a decisive detail, a model must ground its attention on the correct region, but knowing which region is correct presupposes having already observed that detail. We identify this circular dependency as the grounding paradox, show that grounding errors are rarely self-corrected within a single trajectory---once a misleading region is inspected, all subsequent reasoning conditions on that observation and the error propagates to the final answer---and observe that because each trajectory constructs its own evidence, answer-level aggregation discards the very information that distinguishes trajectories. We propose Test-Time Scaling over Perception (TTSP), a closed-loop framework that treats perception as the unit of scalable inference and allocates compute along two axes: Entropy-Gated Perceptual Exploration samples diverse trajectories and uses critical-token entropy to withhold evidence the model cannot commit to, while Evidence-Guided Iterative Refinement distills validated observations into a correctable Evidence Ledger that steers later rounds to re-inspect unresolved regions. Across high-resolution and general multimodal benchmarks, TTSP consistently outperforms strong test-time scaling baselines, while improving grounding quality with favorable token efficiency.

cs.CV↗

arXiv2026

SubFLOT: Submodel Extraction for Efficient and Personalized Federated Learning via Optimal Transport

Federated Learning (FL) enables collaborative model training while preserving data privacy, but its practical deployment is hampered by system and statistical heterogeneity. While federated network pruning offers a path to mitigate these issues, existing methods face a critical dilemma: server-side pruning lacks personalization, whereas client-side pruning is computationally prohibitive for resource-constrained devices. Furthermore, the pruning process itself induces significant parametric divergence among heterogeneous submodels, destabilizing training and hindering global convergence. To address these challenges, we propose SubFLOT, a novel framework for server-side personalized federated pruning. SubFLOT introduces an Optimal Transport-enhanced Pruning (OTP) module that treats historical client models as proxies for local data distributions, formulating the pruning task as a Wasserstein distance minimization problem to generate customized submodels without accessing raw data. Concurrently, to counteract parametric divergence, our Scaling-based Adaptive Regularization (SAR) module adaptively penalizes a submodel's deviation from the global model, with the penalty's strength scaled by the client's pruning rate. Comprehensive experiments demonstrate that SubFLOT consistently and substantially outperforms state-of-the-art methods, underscoring its potential for deploying efficient and personalized models on resource-constrained edge devices.

cs.LG↗

arXiv2024

SoftDedup: an Efficient Data Reweighting Method for Speeding Up Language Model Pre-training

The effectiveness of large language models (LLMs) is often hindered by duplicated data in their extensive pre-training datasets. Current approaches primarily focus on detecting and removing duplicates, which risks the loss of valuable information and neglects the varying degrees of duplication. To address this, we propose a soft deduplication method that maintains dataset integrity while selectively reducing the sampling weight of data with high commonness. Central to our approach is the concept of "data commonness", a metric we introduce to quantify the degree of duplication by measuring the occurrence probabilities of samples using an n-gram model. Empirical analysis shows that this method significantly improves training efficiency, achieving comparable perplexity scores with at least a 26% reduction in required training steps. Additionally, it enhances average few-shot downstream accuracy by 1.77% when trained for an equivalent duration. Importantly, this approach consistently improves performance, even on rigorously deduplicated datasets, indicating its potential to complement existing methods and become a standard pre-training process for LLMs.

cs.CL↗

arXiv2024

Forte: A Suite of Advanced Multireference Quantum Chemistry Methods

Forte is an open-source library specialized in multireference electronic structure theories for molecular systems and the rapid prototyping of new methods. This paper gives an overview of the capabilities of Forte, its software architecture, and examples of applications enabled by the methods it implements.

physics.chem-ph↗

arXiv2023

TeacherLM: Teaching to Fish Rather Than Giving the Fish, Language Modeling Likewise

Large Language Models (LLMs) exhibit impressive reasoning and data augmentation capabilities in various NLP tasks. However, what about small models? In this work, we propose TeacherLM-7.1B, capable of annotating relevant fundamentals, chain of thought, and common mistakes for most NLP samples, which makes annotation more than just an answer, thus allowing other models to learn "why" instead of just "what". The TeacherLM-7.1B model achieved a zero-shot score of 52.3 on MMLU, surpassing most models with over 100B parameters. Even more remarkable is its data augmentation ability. Based on TeacherLM-7.1B, we augmented 58 NLP datasets and taught various student models with different parameters from OPT and BLOOM series in a multi-task setting. The experimental results indicate that the data augmentation provided by TeacherLM has brought significant benefits. We will release the TeacherLM series of models and augmented datasets as open-source.

cs.CL↗

arXiv2023

Sedimentation of a single soluble particle at low Reynolds and high P\'eclet numbers

We investigate experimentally the dissolution of an almost spherical butyramide particle during its sedimentation, in the low Reynolds high P\'eclet regime. The particle sediments in a quiescent aqueous solution, and its shape and position are measured simultaneously by a camera attached to a translation stage. The particle is tracked in real time, and the translation stage moves accordingly to keep the particle in the field of the camera. The measurements from the particle image show that the radius shrinking rate is constant with time, and independent of the initial radius of the particle. We explain this with a simple model, based on the sedimentation law in the Stokes' regime and the mass transfer rate at low Reynolds and high P\'eclet numbers. The theoretical and experimental results are consistent within $20\%$. We introduce two correction factors to take into account the non-sphericity of the particle and the inclusions of air bubbles inside the particle, and reach quantitative agreement. With these corrections, the indirect measurement of the radius shrinking rate deduced from the position measurement is also in agreement with the model. We discuss other correction factors, and explain why there are negligible in the present experiment. We also compute the effective Sherwood number as a function of an effective P\'eclet number.

physics.flu-dyn↗

arXiv2022

Evolution of barchan dune interactions investigated by a downscaled water tunnel experiment: the temporal characteristics and a soliton-like behavior

This paper reports a downscaled water tunnel experiment to study the temporal characteristics of a double dune interaction system and the new pattern of dune interaction when the initial mass ratio of the two dunes is large. These topics are useful for a comprehensive understanding of the dune interaction system but were rarely covered before. The turnover time scale under dune interaction is defined, and its time averaged value is found to have a nonmonotonic relationship with the initial mass ratio. A nonmonotonic relationship is also found between the convexity of the downstream dune tip and the initial mass ratio. The stationary points of the two nonmonotonic curves above correspond to the same dune interaction pattern named 'exchange-chasing', which is considered indispensable in the classification map of dune interactions. The upstream dune acts as an energy transmitter between fluid flow and the downstream dune. A soliton-like behavior occurs when the downstream dune enlarges, where a small dune is detached from the downstream dune tip and gets passed by the upstream dune approximately without mass exchange. The activity of such temporary soliton is found to be negatively related with the initial dune spacing and positively related with the initial mass ratio.

physics.geo-ph↗