arXiv subjects

Guojun Wu

Publications and source records attributed to Guojun Wu.

14 recordsLinked to original sources

arXiv2025

It's Not That Simple. An Analysis of Simple Test-Time Scaling

Prior work proposed simple test-time scaling, a method for replicating this scaling behavior with models distilled from o1-like models by manually controlling test-time compute: either scaling down by enforcing a maximum length or scaling up by iteratively appending "Wait" when the model is about to terminate its generation. This paper presents an analysis of simple test-time scaling and finds that the scaling behavior is largely attributed to scaling down by enforcing a maximum length. In contrast, fine-tuning on long CoT data distilled from o1-like models has no significant impact on scaling behavior, and scaling up by appending "Wait" leads to inconsistencies, as the model may oscillate between solutions. A key distinction exists between scaling down by enforcing a maximum length and scaling up test-time compute in o1-like models, such as DeepSeek-R1\@. These models are typically allowed to utilize as much compute as needed, with the only constraint being the model's maximum supported length. By learning to naturally scale up test-time compute during reinforcement learning, o1-like models surpass their peak performance when scaling up. In contrast, simple test-time scaling progressively imposes a lower upper limit on model performance as it scales down. While replicating the test-time scaling behavior of o1 models can be straightforward by scaling down, it is crucial to recognize that the goal of scaling test-time compute is to unlock higher performance -- beyond what the model could originally achieve -- rather than merely reproducing the appearance of scaling behavior.

cs.LG↗

arXiv2025

Deep Distributional Learning with Non-crossing Quantile Network

In this paper, we introduce a non-crossing quantile (NQ) network for conditional distribution learning. By leveraging non-negative activation functions, the NQ network ensures that the learned distributions remain monotonic, effectively addressing the issue of quantile crossing. Furthermore, the NQ network-based deep distributional learning framework is highly adaptable, applicable to a wide range of applications, from classical non-parametric quantile regression to more advanced tasks such as causal effect estimation and distributional reinforcement learning (RL). We also develop a comprehensive theoretical foundation for the deep NQ estimator and its application to distributional RL, providing an in-depth analysis that demonstrates its effectiveness across these domains. Our experimental results further highlight the robustness and versatility of the NQ network.

stat.ML↗

arXiv2024

The characterization for the sobriety of $L$-convex spaces

With a commutative integral quantale $L$ as the truth value table, this study focuses on the characterizations of the sobriety of stratified $L$-convex spaces, as introduced by Liu and Yue in 2024. It is shown that a stratified sober $L$-convex space $Y$ is a sobrification of a stratified $L$-convex space $X$ if and only if there exists a quasihomeomorphism from $X$ to $Y$; a stratified $L$-convex space is sober if and only if it is a strictly injective object in the category of stratified $S_0$ $L$-convex spaces.

math.CT↗

arXiv2024

Sober $L$-convex spaces and $L$-join-semilattices

With a complete residuated lattice $L$ as the truth value table, we extend the definition of sobriety of classical convex spaces to the framework of $L$-convex spaces. We provide a specific construction for the sobrification of an $L$-convex space, demonstrating that the full subcategory of sober $L$-convex spaces is reflective in the category of $L$-convex spaces with convexity-preserving mappings. Additionally, we introduce the concept of Scott $L$-convex structures on $L$-ordered sets. As an application of this type of sobriety, we obtain a characterization for the $L$-join-semilattice completion of an $L$-ordered set: an $L$-ordered set $Q$ is an $L$-join-semilattice completion of an $L$-ordered set $P$ if and only if the Scott $L$-convex space $(Q, σ^{\ast}(Q))$ is a sobrification of the Scott $L$-convex space $(P, σ^{\ast}(P))$.

math.GN↗

arXiv2024

Topological representations for frame-valued domains via $L$-sobriety

With a frame $L$ as the truth value table, we study the topological representations for frame-valued domains. We introduce the notions of locally super-compact $L$-topological space and strong locally super-compact $L$-topological space. Using these concepts, continuous $L$-dcpos and algebraic $L$-dcpos are successfully represented via $L$-sobriety. By means of Scott $L$-topology and specialization $L$-order, we establish a categorical isomorphism between the category of the continuous (resp., algebraic) $L$-dcpos with Scott continuous maps and that of the locally super-compact (resp., strong locally super-compact) $L$-sober spaces with continuous maps. As an application, for a continuous $L$-poset $P$, we obtain a categorical isomorphism between the category of directed completions of $P$ with Scott continuous maps and that of the $L$-sobrifications of $(P, σ_{L}(P))$ with continuous maps.

math.GN↗

arXiv2024

Representations of FS-domains and BF-domains via FS-approximation Spaces

In this paper, concepts of (topological) FS-approximation spaces are introduced. Representations of FS-domains and BF-domains via (topological) FS-approximation spaces are considered. It is proved that the collection of CF-closed sets in an FS-approximation space (resp., a topological FS-approximation space) endowed with the set-inclusion order is an FS-domain (resp., a BF-domain) and that every FS-domain (resp., BF-domain) is order isomorphic to the collection of CF-closed sets of some FS-approximation space (resp., topological FS-approximation space) endowed with the set-inclusion order. The concept of topological BF-approximation spaces is introduced and a skillful method without using CF-approximable relations to represent BF-domains is given. It is also proved that the category of FS-domains (resp., BF-domains) with Scott continuous maps as morphisms is equivalent to that of FS-approximation spaces (resp., topological FS-approximation spaces) with CF-approximable relations as morphisms.

math.CT↗

arXiv2024

When do CF-approximation spaces capture sL-domains

In this paper, by means of upper approximation operators in rough set theory, we study representations for sL-domains and its special subclasses. We introduce the concepts of sL-approximation spaces, L-approximation spaces and bc-approximation spaces, which are special types of CF-approximation spaces. We prove that the collection of CF-closed sets in an sL-approximation space (resp., an L-approximation space, a bc-approximation space) ordered by set-theoretic inclusion is an sL-domain (resp., an L-domain, a bc-domain); conversely, every sL-domain (resp., L-domain, bc-domain) is order-isomorphic to the collection of CF-closed sets of an sL-approximation space (resp., an L-approximation space, a bc-approximation space). Consequently, we establish an equivalence between the category of sL-domains (resp., L-domains) with Scott continuous mappings and that of sL-approximation spaces (resp., L-approximation spaces) with CF-approximable relations.

math.CT↗

arXiv2024

Evaluating Automatic Metrics with Incremental Machine Translation Systems

We introduce a dataset comprising commercial machine translations, gathered weekly over six years across 12 translation directions. Since human A/B testing is commonly used, we assume commercial systems improve over time, which enables us to evaluate machine translation (MT) metrics based on their preference for more recent translations. Our study not only confirms several prior findings, such as the advantage of neural metrics over non-neural ones, but also explores the debated issue of how MT quality affects metric reliability--an investigation that smaller datasets in previous research could not sufficiently explore. Overall, our research demonstrates the dataset's value as a testbed for metric evaluation. We release our code at https://github.com/gjwubyron/Evo

cs.CL↗

arXiv2024

Representations of domains via closure spaces in the quantale-valued setting

With a commutative unital quantale $L$ as the truth value table, this study focuses on the representations of $L$-domains by means of $L$-closure spaces. First, the notions of interpolative generalized $L$-closure spaces and directed closed sets are introduced. It is proved that in an interpolative generalized $L$-closure space (resp., $L$-closure space), the collection of directed closed sets with respect to the inclusion $L$-order forms a continuous $L$-dcpo (resp., an algebraic $L$-dcpo). Conversely, it is shown that every continuous $L$-dcpo (resp., algebraic $L$-dcpo) can be reconstructed by an interpolative generalized $L$-closure space (resp., $L$-closure space). Second, when $L$ is integral, the notion of dense subspaces of generalized $L$-closure spaces is introduced. By means of dense subspaces, an alternative representation for algebraic $L$-dcpos is given. Moreover, the concept of $L$-approximable relations between interpolative generalized $L$-closure spaces is introduced. Consequently, a categorical equivalence between the category of interpolative generalized $L$-closure spaces (resp., $L$-closure spaces) with $L$-approximable relations and that of continuous $L$-dcpos (resp., algebraic $L$-dcpos) with Scott continuous mappings is established.

math.GN↗

arXiv2024

StreakNet-Arch: An Anti-scattering Network-based Architecture for Underwater Carrier LiDAR-Radar Imaging

In this paper, we introduce StreakNet-Arch, a real-time, end-to-end binary-classification framework based on our self-developed Underwater Carrier LiDAR-Radar (UCLR) that embeds Self-Attention and our novel Double Branch Cross Attention (DBC-Attention) to enhance scatter suppression. Under controlled water tank validation conditions, StreakNet-Arch with Self-Attention or DBC-Attention outperforms traditional bandpass filtering and achieves higher $F_1$ scores than learning-based MP networks and CNNs at comparable model size and complexity. Real-time benchmarks on an NVIDIA RTX 3060 show a constant Average Imaging Time (54 to 84 ms) regardless of frame count, versus a linear increase (58 to 1,257 ms) for conventional methods. To facilitate further research, we contribute a publicly available streak-tube camera image dataset contains 2,695,168 real-world underwater 3D point cloud data. More importantly, we validate our UCLR system in a South China Sea trial, reaching an error of 46mm for 3D target at 1,000 m depth and 20 m range. Source code and data are available at https://github.com/BestAnHongjun/StreakNet .

cs.CV↗

arXiv2024

ICU: Conquering Language Barriers in Vision-and-Language Modeling by Dividing the Tasks into Image Captioning and Language Understanding

Most multilingual vision-and-language (V&L) research aims to accomplish multilingual and multimodal capabilities within one model. However, the scarcity of multilingual captions for images has hindered the development. To overcome this obstacle, we propose ICU, Image Caption Understanding, which divides a V&L task into two stages: a V&L model performs image captioning in English, and a multilingual language model (mLM), in turn, takes the caption as the alt text and performs cross-lingual language understanding. The burden of multilingual processing is lifted off V&L model and placed on mLM. Since the multilingual text data is relatively of higher abundance and quality, ICU can facilitate the conquering of language barriers for V&L models. In experiments on two tasks across 9 languages in the IGLUE benchmark, we show that ICU can achieve new state-of-the-art results for five languages, and comparable results for the rest.

cs.CL↗

arXiv2023

Representations of Domains via CF-approximation Spaces

Representations of domains mean in a general way representing a domain as a suitable family endowed with set-inclusion order of some mathematical structures. In this paper, representations of domains via CF-approximation spaces are considered. Concepts of CF-approximation spaces and CF-closed sets are introduced. It is proved that the family of CF-closed sets in a CF-approximation space endowed with set-inclusion order is a continuous domain and that every continuous domain is isomorphic to the family of CF-closed sets of some CF-approximation space endowed with set-inclusion order. The concept of CF-approximable relations is introduced using a categorical approach, which later facilitates the proof that the category of CF-approximation spaces and CF-approximable relations is equivalent to that of continuous domains and Scott continuous maps.

math.RA↗

arXiv2021

Rating Facts under Coarse-to-fine Regimes

The rise of manipulating fake news as a political weapon has become a global concern and highlighted the incapability of manually fact checking against rapidly produced fake news. Thus, statistical approaches are required if we are to address this problem efficiently. The shortage of publicly available datasets is one major bottleneck of automated fact checking. To remedy this, we collected 24K manually rated statements from PolitiFact. The class values exhibit a natural order with respect to truthfulness as shown in Table 1. Thus, our task represents a twist from standard classification, due to the various degrees of similarity between classes. To investigate this, we defined coarse-to-fine classification regimes, which presents new challenge for classification. To address this, we propose BERT-based models. After training, class similarity is sensible over the multi-class datasets, especially in the fine-grained one. Under all the regimes, BERT achieves state of the art, while the additional layers provide insignificant improvement.

cs.CL↗

arXiv2019

Reward Advancement: Transforming Policy under Maximum Causal Entropy Principle

Many real-world human behaviors can be characterized as a sequential decision making processes, such as urban travelers choices of transport modes and routes (Wu et al. 2017). Differing from choices controlled by machines, which in general follows perfect rationality to adopt the policy with the highest reward, studies have revealed that human agents make sub-optimal decisions under bounded rationality (Tao, Rohde, and Corcoran 2014). Such behaviors can be modeled using maximum causal entropy (MCE) principle (Ziebart 2010). In this paper, we define and investigate a general reward trans-formation problem (namely, reward advancement): Recovering the range of additional reward functions that transform the agent's policy from original policy to a predefined target policy under MCE principle. We show that given an MDP and a target policy, there are infinite many additional reward functions that can achieve the desired policy transformation. Moreover, we propose an algorithm to further extract the additional rewards with minimum "cost" to implement the policy transformation.

cs.AI↗